Start with the Marshall Attack, because it is the clearest case of what these players are fleeing.

Across 2015–2023, 72% of correspondence games between players both rated above 2400 that reached the Marshall — 336 of 464 — got to move 20 without either side playing a move that had not appeared in an earlier game in the database. The median such game does not diverge from the record until deep in the twenties. The Marshall in correspondence has become a recital: a drawing mechanism whose main lines are mapped, engine-checked, and available to both players before the first move. To enter it is to agree to play twenty moves of other people’s chess.

Openings like that are what this piece is about — and, more precisely, what the elite repertoire does to avoid them. There is a common story about the engine era in which opening theory gradually fills up until the game is solved from the first move. The correspondence data tell a stranger story. Theory is filling up. But at the top, the games are staying ahead of it, and the way that happens is visible in the numbers.

One scoping point governs everything after. What I measure is the move at which a game departs from the public correspondence record — the 1.56 million prior rated games in this corpus. That is not where the players stopped relying on preparation; they prepare from engines, from over-the-board games, from private files that never reach a rated game. A move new to this database at move 14 is often not new to the two players. So this is a repetition rate for the shared correspondence book, not a measure of how much chess has been solved. Every figure is a floor.

The flat line that wasn’t

Take the obvious measure first: the share of elite games that reach move 20 with no new move, year by year. Plot it and you get something that looks like a plateau. For players both above 2400, the share climbed from around 2% in the late 1990s to roughly 17% by the mid-2010s — and then, from 2015 to 2023, it holds near 17%, barely moving. The genuine top of the game, players both above 2500, shows the same shape one notch lower, flat near 14%.

For a while I believed that flat stretch was the finding: theory approaching saturation, the book filling until it could fill no faster. It is not. The flatness is an average concealing two large and opposite movements — theory deepening within openings, and the repertoire shifting toward the openings where it deepened least — and separating them is the actual story.

Hold the opening repertoire fixed — freeze each opening’s share of elite games at its 2010–2014 level, then apply the bookedness each opening actually had in later years. The flat line comes apart. On a fixed repertoire, elite bookedness does not plateau after 2015; it climbs from about 15% to about 22%. Theory kept deepening inside the openings. What flattened the aggregate was that the elite repertoire moved toward the openings where it was deepening least.

The decomposition is clean, and it points the other way from the naive reading. Of the aggregate change from 2010–2014 to 2020–2023, deepening within openings contributes +4.4 percentage points; the shift between openings contributes −2.6. The two nearly cancel, and their near-cancellation is the flat line. The within figure is an upper bound — the openings players kept are, by the same logic, selected for slower deepening, so a truly constant repertoire would have deepened somewhat less than +4.4pp — but the direction is not in doubt, and it is robust: the split holds between +4.1/−2.4 and +4.6/−3.0 across base periods and after dropping the smallest openings. Crucially, the deepening is concentrated in the sharp, linear openings, not the flexible ones: the Semi-Slav booked up +12pp, the Grünfeld +14pp, the French +9pp over the decade, while the flexible, transposition-prone systems (English, Réti, Italian) contributed far less to the within-opening rise. Whatever the exact magnitude, the book did not stop filling; the repertoire shifted out from under it.

Where they ran from, and to

The direction of the flight is exactly what the flat aggregate hides, and it is not random.

Rank the elite openings of 2015–2023 by how booked they are. At the top, the sharp, heavily analysed lines: the Marshall at 72%, the Semi-Slav at 31%, the Najdorf at 30%, the Rossolimo at 28%, the Grünfeld at 27%. At the bottom, the openings where independent play still begins early: the London systems at 1–2%, the Italian at 3%, the Réti at 4%, the English at 6%, the Queen’s Gambit Declined at 8%.

Now set that against the repertoire shifts I documented in the earlier piece on opening popularity. The openings the elite has been shedding are the booked ones. The openings it has adopted are the unbooked ones. The migration runs down this table — out of the Najdorf and Semi-Slav at thirty per cent booked, into the Italian and English at three and six. The players moving to the Italian were not choosing it for its ideas; they were choosing it for its emptiness.

And the migration is not merely contemporaneous with the book closing — it lags it, in the direction cause would run. An opening’s bookedness measured in 2010–2014 is associated with how much elite share it subsequently gained or lost over the following decade, at a rank correlation of −0.74. The openings already heavily booked in the early 2010s are the ones players left afterward. The predictor sits entirely before the migration window, so the relationship owes nothing to the recent data. This is an association, not a proven mechanism — a common cause could drive both, if sharp, narrow lines simply attract early heavy analysis and later abandonment for their own reasons. But the ordering is at least consistent with the book closing first and the players leaving after.

Two ways an opening empties — offered as hypothesis

Look at which openings sit furthest from that trend and a rough split appears. Some lost elite share after becoming heavily booked — the Najdorf, the Semi-Slav — the migration this piece is about, players leaving once the theory was exhausted. Others lost share while barely booked at all: the King’s Indian (about 9% booked) and the French (about 14%) declined despite unmapped territory remaining. These do not fit that story — there was no full book to leave — and I flag them as the exception rather than fold them in: their decline is consistent with the explored fraction having already returned a verdict bad enough that mapping the rest was beside the point.

I want to be exact about the status of this split. It is read off the residuals of a one-factor relationship — a hypothesis, not a result — and the same facts admit a simpler reading with no two mechanisms at all: sharp, narrow openings may attract both early heavy analysis and later abandonment as a single tendency, with no separate “exhaustion” and “verdict” stories needed. Distinguishing these would require bringing engine evaluations of the resulting positions into the model, which this analysis does not do. The Petroff is the case that keeps the question open: heavily booked and yet rising in elite share, because when the booked lines are the safe, near-symmetrical positions elite players want as Black, a closed book is a feature rather than something to flee.

A note before the recent data: I stop at 2023

The 2024 and 2025 figures break the pattern — the raw 2400+ share jumps to 22–24%. I do not believe that jump reflects a change in play, and the analysis above deliberately ends at 2023. The short version: almost all of the surge is a batch of games in one narrow rating band entering through a database update, not more or deeper chess. It is real enough in the numbers to need explaining and contaminated enough that its size cannot be trusted. The full diagnosis, and an open question I would like readers’ help with, is in the notes.

What is left

The sealed move — the move a correspondence player settles on alone, the act this publication is named for — arrives, in the shared record, later than it used to: around move 14 in the average elite game, past move 20 in the Marshall. But the deepening those numbers understate is not where the top players spend most of their games, because their repertoire has drifted elsewhere on the board — into the Italian, the English, the slow queen’s-pawn systems — where the record still runs out near move 9 and the game is still theirs to make.

That is the shape the numbers take in the engine era, at least in correspondence: within openings, the public book keeps deepening; in aggregate, the elite repertoire drifts toward the openings where it has deepened least. Whether that drift is deliberate strategy, a thousand independent choices that happen to share a direction, or openings simply following the same curve at different stages, the data cannot say. The underlying role of preparation is certainly deeper than this public-record measure can show. What the measure does show is a repertoire in steady motion down the bookedness gradient — the strongest players, game by game, ending up in the openings where the record still runs out early.

SIM Paweł Fiedor - The Sealed Move

Notes on method and open questions

The corpus and the measure. 1.56 million rated correspondence games through May 2026 (Opening Master OM CORR). “Novelty” is the first ply at which a game’s move sequence departs from all earlier-dated games in the corpus (same-date ties by file order; unknown month/day placed mid-year). Move-sequence hashing to a 60-ply cap. Chess960 and games under 15 full moves excluded. Cohorts require both players above the stated rating; “field” is all other rated games. Annual 2400+ samples run 3,854–6,980 in the 2015–2023 window; 2500+ runs 400–1,340.

The shift-share decomposition. Aggregate elite bookedness changed +2.0pp from 2010–2014 to 2020–2023. Decomposing at fixed 2010–2014 opening weights: within-opening deepening +4.4pp, between-opening repertoire shift −2.6pp, interaction +0.2pp. Holding weights at 2010–2014, the annual series rises from ~15% (2013–14) to ~22% (2022–23) rather than holding flat. The near-flat raw aggregate is therefore compositional, not a ceiling on theory. Robustness: the split holds at +4.6/−3.0 (2008–12 base), +4.4/−2.6 (2010–14), +4.1/−2.4 (2012–16), and +4.2/−2.4 after excluding openings with fewer than 500 base-period games — so it is not an artifact of the base period or the small-opening tail. Identification caveat: the +4.4pp within figure is an upper bound. The fixed-weight counterfactual assumes each opening’s bookedness path is independent of whether players kept playing it, which is not credible — players abandon the fastest-deepening openings, so the retained repertoire is selected toward slower deepening, and a genuinely constant repertoire would have deepened by somewhat less. The sign and rough magnitude survive (the deepening is large and concentrated in openings players did keep, such as the Najdorf and Ruy), but the figure should be read as “at most about +4.4pp,” not a point estimate.

Migrating-in vs migrating-out bookedness. The openings gaining elite share are much less booked than those losing it (2015–2023): English 5.7%, QGD 7.7%, Italian 3.3%, London 2.4% — against Najdorf 30.1%, Semi-Slav 31.1%, French 13.9%, King’s Indian 9.3%. The weight shift is thus from high-bookedness to low-bookedness groups, which is what pulls the aggregate down even as each group deepens.

Transposition, and where the deepening lives. A line-based measure reads a known position reached by a new move-order as a false novelty, and the elite has shifted toward transposition-heavy systems (English, Réti, Catalan) — so one might worry the within-opening deepening is a move-order artifact concentrated in those flexible webs. It is not. Splitting the +4.4pp within-component by opening type, the linear, forcing openings (where transposition cannot manufacture novelties) contribute about +2.0pp and the flexible systems only about +0.7pp; the sharpest openings show the largest within-opening rises (Semi-Slav +12pp, Grünfeld +14pp, French +9pp, 2010–14 to 2020–23), while the flexible group deepened far less in absolute terms. So the deepening is driven by openings immune to the transposition bias, not by it. The migration into flexible systems is also modest (8.7% → 11.9% of elite games). The line-based measure biases the level down — figures are floors — but does not create the decomposition.

Over-the-board theory — a two-sided bias. The corpus is correspondence-only. This creates two opposing biases on the recent within-opening series, and I cannot cleanly size either. On one side, correspondence players import over-the-board novelties — increasingly from the neural-network era — and those register as correspondence novelties because they never appeared in a rated correspondence game, which deflates measured bookedness more in the later period. On the other, players may follow well-mapped OTB and engine lines further precisely because those lines are now safer and better analysed, which inflates measured bookedness in recent years. Both channels are present; I have no evidence on their relative magnitude, so I do not claim the true within-opening rise is necessarily steeper than measured — only that the level is a floor and the recent trend carries this unresolved two-sided uncertainty. Testing it would require checking positions against an over-the-board database, which is future work.

Is the raw aggregate “flat”? Regressing raw annual bookedness on year over 2015–2023: 2500+ slope −0.09 pp/yr (p = 0.48); 2400+ slope +0.21 pp/yr (p ≈ 0.05), about +1.9pp across the window. That is not noise to wave away — it matches the +2.0pp aggregate change the decomposition reports. The point is that this slow raw rise is a mirage: it masks a much larger within-opening deepening offset by a flight toward less-booked openings. “Flat vs slightly rising” in the raw series is the wrong question; the real motion is the split beneath it.

The lagged relationship. 2010–2014 bookedness vs 2010–2014 → 2021–2025 share change, 20 opening groups with n ≥ 300, Spearman ρ = −0.74. Leave-one-out holds it between −0.70 and −0.81. Bookedness is uncorrelated with an opening’s raw game count (r = −0.01); ranking stable across windows (ρ = 0.92). Chart title states association, not cause, deliberately.

The Marshall figure. 72.4% on n = 464; adding or removing five booked games moves it about a point (71.3–73.5%).

The 2024–25 break — and a question. The recent jump is not spread across correspondence chess. Total rated volume barely moved (about 46,000 games in 2023, 59,000 in 2025). But by the lower-rated player’s band, one slice tripled: games with the lower player in the 2400–2499 band went from about 3,600 to over 11,000, while every other band held flat or fell — one (2000–2199) shrank. That single-band flood is the signature of a batch entering the database, in my case the May 2026 OM CORR, not a change in play; because the elite cohort needs both players above 2400, tripling its entry band reshapes the pool. Two things complicate a clean “ignore it”: the new 2400–2499 games are individually more booked than the same band in 2021 (by about six points), and the 2500+ cohort, which no batch inflated, also rose over the two years (about 14% to 19%). The higher within-band bookedness is consistent with a non-random selection of events or players — for instance a rating-normalisation patch bumping historical, heavily-booked 2350-rated games into the 2400 bracket, a common cause of single-band floods in third-party database updates — but the current metadata do not let me test that. Uniform downsampling cannot settle it: thinning a batch-dominated pool keeps the batch, which is why the band decomposition, not the downsampling, is the honest test. So the open question, for a readership that knows this data: what did the May 2026 OM CORR actually add — which tournament cycle, which server backlog, or which rating adjustment? If you know, you can probably tell me whether the last two years are noise or news.

First, an apology. The gap since the last post is longer than I intended, and I won’t dress it up as anything other than life getting in the way. Thank you to everyone who stayed subscribed through the silence.

Second, to everyone playing in the Champions League: good luck. The original group allocations were scrapped and the redrawn groups only appeared today, six days before play begins on the 15th. If you spent the past fortnight preparing against opponents you no longer face, you have my sympathy. May your first moves arrive on time even if the pairings didn’t.

While I was away from the keyboard I was not away from the database. I have a correspondence database running up to May 2026, and I wanted an answer to a simple question: what do strong correspondence players actually open with, and how has that changed over the engine era? Counts, not impressions.

The setup

The raw material is roughly 1.6 million rated correspondence games. I split them into two groups. The first contains games where both players were rated above 2400. I will call this the 2400+ group rather than “the elite”: in modern ICCF terms it spans everyone from strong masters to the handful of players who actually steer the metagame, and a sceptic could reasonably ask for the same cut at 2500. I ran it. At 2500 the annual samples are small (roughly 300–1,700 games), so only period averages are usable, but on those averages every gainer below still gains and every loser still loses, several of the declines are steeper, and the one sign that flips concerns the Najdorf — covered in its own section, where the flip turns out to strengthen the argument rather than weaken it. The second group is everyone else with a rating. Unrated games were dropped.

Annual sample sizes for the 2400+ group run from roughly 1,900 games in 2000 to about 7,000 around 2021, dip to ~4,000 in 2022–23, then jump to ~10,000–12,000 in 2024–25 as recent events enter the database. The wider field contributes 23,000–84,000 games a year. Those swings are why every figure below is a within-group share for its year, why the small-multiples panels use a three-year average, and why I will not lean on any single recent year. The Najdorf chart is the deliberate exception: it is shown unsmoothed, with the coverage artifacts visible and flagged, because its 2015 peak is a specific annual value and I would rather show the noise than hide it. The partial year 2026 is excluded.

Every game was reclassified against a single ECO reference, so the trend lines measure changes in what people play rather than changes in how databases were tagged. ECO classification has a known weakness this audience will spot immediately: transposition. Move orders designed to dodge preparation blur the English into the QGD and Catalan complex, and some of my buckets (”Rossolimo / misc”, “Queen’s Pawn / London”) are visibly heterogeneous. The headline shifts are far too large to be classification noise, but treat single-point differences accordingly. One more caveat in the same spirit: the 2400+ population itself is not constant over 25 years. Changes in who plays serious correspondence chess could move these lines independently of anything engines say. I can control for tagging drift; I cannot control for player-pool drift.

What moved

Eight openings, twenty-five years, two rating groups. The panels were chosen as the largest shares and largest movers; the full table covers 47 opening groups and ships with the follow-up post.

The gainers first, described as the lines actually run. The Queen’s Gambit Declined roughly tripled its 2400+ share since the early 2000s and is still rising. The Italian, a rounding error above 2400 in 2004, now takes over 3% of games. The English is a recovery story rather than a steady climb: it sagged from around 5% to under 4% by the mid-2010s, then rose to over 7%. The Semi-Slav is the awkward one for any tidy narrative, and I’d rather flag it than smooth it: it surged from under 3% to a peak near 9% around 2013–14 and has been retreating for a decade since, to between 6 and 7%. Whatever the 2400+ group found in the Semi-Slav in the late 2000s, it has been slowly walking away from it ever since, while the wider field’s share is still near its high.

The losers are more uniform. The French fell at the top from over 7% of games to 2%, a steeper fall than in the wider field. The King’s Indian followed the same path. The sharp Sicilians thinned out one by one, and the Scheveningen, an opening Kasparov built a world championship career on, went from 1.9% of 2400+ games to 0.1% — a few dozen games across the last five years, out of some 37,000. I called that functionally extinct in an earlier draft and I’ll stand by it; at this sample size the share could double and the conclusion wouldn’t move.

Then versus now

The slope chart compares 2000–04 with 2021–25. Endpoint comparisons hide timing, as the Semi-Slav above and the Najdorf below both demonstrate, so read this chart for magnitudes and direction, then check the time series before trusting any individual line.

One line justifies the whole exercise. In the first version of this analysis the Ruy Lopez appeared as a single block and looked like a modest gainer. Splitting out the Berlin complex (C65–C67, which includes the 4.d3 Anti-Berlin, so not every game here is an endgame grind) changes the picture: the Berlin went from 0.4% of 2400+ games to 5.0%, the steepest rise of any line measured, while the rest of the Spanish declined from 6.9% to 4.9%. The Ruy Lopez did not grow. The Berlin ate it from the inside, and the aggregate hid the meal.

The timing is worth a sentence, because it complicates the story I wanted to tell. The Berlin’s rise is concentrated between 2010 and 2015 — from 1% to over 4% — and it has plateaued since. That is before the NNUE era, and it coincides with the opening’s over-the-board revival. Correspondence players reached the same verdict as Kramnik in 2000 with rather more computation behind them, but on this evidence they reached it when the OTB elite did, not years ahead. The diffusion ran in both directions.

Two reading aids for the chart as a whole. The bolded gold and grey lines mark shifts above 1.5 percentage points. And despite all the crossing lines, the top of the distribution remains concentrated: the five largest openings still account for roughly two-fifths of 2400+ games. The repertoire narrowed; it did not scatter.

The Najdorf

The Najdorf is where the slope chart and the time series appear to disagree, and the disagreement is instructive. On the slope chart the Najdorf gained: 11.3% in 2000–04 to 12.3% in 2021–25. On the time series it peaked at 17.5% in 2015 and has eroded since, to around 12% in recent years. Both are true. The slope chart’s endpoints straddle the peak, so a rise-and-decline registers as a small net gain. This is the strongest argument for publishing the time series alongside any before/after comparison.

So the accurate statement is double-edged. The Najdorf has lost roughly a third of its peak share among 2400+ players over the past decade. It also remains the most played opening in that group by a wide margin, and the wider field has kept adopting it throughout, from about 5% in the early 2000s to between 9 and 10% now.

The 2500 cut sharpens this picture. For pairs both rated above 2500, the endpoint comparison itself turns negative: 14.3% in 2000–04 down to 12.5% in 2021–25. The Najdorf was already near its ceiling at the very top of the rating list two decades ago, and the higher the band, the earlier the exit begins. That gives the chart a third rung. The 2500+ group peaked first, the 2400 band peaked around 2015, and the field below is still climbing. Whatever is driving the migration moves down the rating list from the top.

Why the divergence between the groups? I see two candidate explanations, and the data cannot separate them.

The first is discovery exhaustion. Twenty years of engine analysis at correspondence time controls is a discovery process whose scale, if not its kind, has no real OTB equivalent, and the 2400+ group ran it first and hardest. As viable novelties thin out at the depths they analyse, the opening’s practical yield above 2400 falls, and the share follows with it. On this reading the wider field plays the Najdorf today for the reason the 2400+ group played it in 2010, and the lag is informational.

The second is incentives. Above 2400, both sides reach the same engine-approved positions and implement them near-perfectly; the marginal value of sharpness is close to zero, so players drift toward lines that are cheaper to maintain. Below 2400, opponents’ implementation is leakier and sharpness still converts into points. On this reading the lag is not the field catching up to the elite; it is two populations rationally choosing differently because they face different opponents. Nothing in these charts distinguishes the two stories, and the honest position is that both are probably operating.

And why does the Najdorf keep the crown despite the erosion? Because nothing refuted it. My reading — and this is a conjecture about engine behaviour, not something these charts measure — is that its critical evaluations hold at depth in a way the Scheveningen’s did not: engine-stable sharpness rather than a tactical lottery. Testing that properly would mean line-level evaluation tracking across engine generations, which is a different article. The migration in these charts is not from sharp openings to quiet ones. It is from lines whose evaluations wobble under depth to lines whose evaluations hold, and the Najdorf, inconveniently for a tidy headline, sits in the second group.

What I take from this

The demonstrated finding is descriptive: since 2000, 2400+ correspondence players have concentrated their repertoire into a narrower set of openings whose engine evaluations are stable at extreme depth, and they have moved measurably faster than the rest of the rated field. The mechanism — verdict-driven abandonment, incentive differences, pool composition, or all three — is interpretation, and I have tried to mark the boundary honestly.

For repertoire purposes the practical readings survive the caveats. The 2400+ group’s choices have tended to move ahead of the wider field in these charts, and on that record they are a useful early signal for where the rest of the rated pool is heading — not a validated forecast, and the Berlin episode is a warning that the signal sometimes arrives from outside correspondence entirely. And for readers who love the Scheveningen, the data’s advice is unsentimental: play it below 2400, where it still exists.

The closing thought connects to the engine-philosophy material from earlier posts, stated at the strength the evidence supports. These charts show selection toward openings that survive deep scrutiny. They do not show that correspondence chess has found the truth of the opening; they show what remains when twenty-five years of computation strips away everything that doesn’t hold. Whether that residue is solid ground or just the ground nobody has yet found the fault line in is a question for a future database.

If there’s interest, the full pipeline behind these charts, from raw PGN to the final images, will be the subject of a follow-up post for anyone who wants to run the same analysis on their own database, at either rating cut.

SIM Paweł Fiedor

Source: correspondence database to May 2026, ~1.6M rated games. 2400+ group: both players rated above 2400 (annual n ≈ 1,900–12,000); wider field: all other rated games (annual n ≈ 23,000–84,000). Sensitivity cut at 2500 (annual n ≈ 300–1,700, period averages only): directions of change unchanged for every group except the Najdorf, whose small endpoint gain at 2400 becomes a decline at 2500 — see the Najdorf section. Endpoint shares (2000–04 → 2021–25, %):

Correspondence chess is, at its core, a research game. The time controls are generous by design — not to let you procrastinate, but to let you work. And working well means having the right tools. Over the years I’ve settled into a fairly stable toolkit, and I’m often asked what it looks like. This post is the answer.

I’ll go through five categories: opening books in the original sense of the word, databases, chess software, engines, and finally endgame resources — both tablebases and the more exotic tools for positions that fall beyond their reach.

Physical Books — Still Alive, and Still Dangerous

There’s a reflex in the correspondence chess community to dismiss printed theory as obsolete the moment a new engine analysis session can be run. This is a mistake, in two distinct directions.

The first is obvious: a well-written theoretical monograph compresses accumulated human understanding in a form that no database search replicates. A book on the Sicilian Sveshnikov or the Nimzo-Indian doesn’t just tell you which moves have been played — it tells you why, which structures are dangerous to both sides, and which endgames to steer toward or avoid. Volumes from Quality Chess, New In Chess, and ChessBase’s own FritzTrainer series give your engine analysis a strategic framework to work within. Raw engine output without that context can lead you down technically sound but practically confused lines.

The second point is one I enjoy more: your opponent is probably using books too, and books contain errors. A serious correspondence player who buys a major theoretical work — say, Kotronias’s treatment of the Sicilian Sveshnikov — doesn’t just read it for the ideas. They stress-test its key assessments against current engines. Sometimes they find a mistake the author missed. And if you know your opponent is working from the same book, and you’ve found that mistake, and you play into the position where it occurs... the game can effectively be over before your opponent realises anything has gone wrong. This is not hypothetical: it’s a real preparation method with a real track record. Correspondence World Champion Jon Edwards described exactly this approach in a Perpetual Chess Podcast interview — he bought Kotronias’s Sveshnikov book, found an error, and subsequently got an opponent to walk into the mistake in a game. Whether that game was correspondence or over the board he left pleasingly ambiguous, but the principle applies equally either way. The book you buy to study from is also, if you’re thorough enough, a map of the traps your opponents may be walking toward.

This works both ways. Your own theoretical literature should be verified regularly. When you rely on a source, run its critical assessments through your current engine setup. The speed of modern engine work means that analysis published even five years ago may have been overtaken — not because the authors were careless, but because Stockfish 10 or an early Lc0 network is simply not the same animal as what’s available now. Trusting a source you haven’t checked is trusting a ghost.

The practical upshot: keep your core theoretical books. Maintain awareness of which authors your opponents favour — their game history usually makes this transparent. Verify the critical lines, and occasionally look for the cracks.

Opening Databases

In correspondence chess, preparation at the board begins with preparation in the database. Before you sit down to analyse, you want to know what roads have already been walked — and how far, and by whom.

There are three sources worth having, each doing something the others don’t.

ChessBase Mega + Corr

The Mega Database 2026 (€229.90) holds over 11.7 million games spanning 1475 to 2025, with more than 114,000 annotated games — the annotated portion alone is a substantial reason to own it. It comes with a weekly update service adding roughly 5,000 new games through the end of 2026, meaning approximately 250,000 additional games by December. This is your OTB anchor: Candidates, Olympiads, World Championships, all the classical events that define opening theory.

For correspondence chess specifically, the Corr Database 2026 (€199.90) is arguably more targeted. It covers 1804 to 2025 with over 2.5 million correspondence games across more than 80,000 tournaments. Correspondence games, played with engines and virtually unlimited thinking time, frequently establish theory months or years before it appears in OTB play. If you want to know what the sharpest theoretical lines look like under serious engine preparation, the Corr database is where you find out.

If you’re building from scratch, the ChessBase 26 Premium Package (€499.90) bundles the program, Mega, Corr, a year of Premium account, six issues of ChessBase Magazine, and 1,000 Ducats. Given that Mega and Corr alone list at nearly €430, the bundling makes sense if you don’t already own a recent ChessBase version.

Opening Master (OM OTB + OM CORR)

Opening Master takes a different approach. OM OTB runs to over 10.5 million OTB games updated monthly, adding roughly 100,000 games per update, and is available via annual subscription at approximately €59. OM CORR holds 2.3 million correspondence games, is widely used in the ICCF community — Opening Master is an ICCF database sponsor — and is available separately at approximately €39 per year. Both databases are delivered in CBV and PGN formats, making them compatible with ChessBase, Chess Assistant, and most other software environments.

The philosophical difference from ChessBase is worth noting: Opening Master is subscription-based, leaner, and constantly updated rather than sold as a yearly snapshot. What you lose in annotation depth you gain in currency. One practical tip for correspondence use: don’t work with OM CORR in its entirety when preparing in a specific opening. Filter it down to the top Elo bracket — games from the strongest correspondence players only — and you get a much sharper signal about where current high-level theory actually stands, without the noise of games from lower levels where engine preparation is less rigorous. The same applies to OM OTB when you want to understand what elite OTB theory looks like in a given structure. Many serious correspondence players run both Opening Master and ChessBase databases in parallel, treating them as complementary rather than competing sources.

It’s worth noting that Chessify — covered in more detail in the software section — also bundles database access into its higher plans: the Lichess game database, a large OTB games collection comparable in scope to Mega, and the ICCF correspondence database of over two million games. If you’re already subscribing to Chessify for engine access, these come included and are accessible directly in the browser without any local setup. They won’t replace dedicated local databases for deep preparation work, but they’re a convenient addition that makes Chessify more of an all-in-one platform than it might first appear.

ChessDB (chessdb.cn)

A free resource that belongs in every serious player’s toolkit: the Chinese Chess Cloud Database at chessdb.cn is a continuously updated database of engine-analysed positions built through a distributed computing project. Where Mega and OM CORR give you historical human and correspondence games, ChessDB gives you something different — positions that have been explored by strong engines to significant depth, with move scores and principal variations attached. The practical use in correspondence chess is looking up a specific position and seeing not just what has been played, but what engines have collectively concluded about the main lines, often going considerably deeper than any single analysis session on your own hardware would produce. It’s free, browser-based, and has an API for programmatic access. Not a replacement for local databases, but a genuinely useful sanity check and discovery tool, particularly when establishing whether a line you’ve been investigating has already been resolved in the broader engine community.

Chess Software: ChessBase vs. Aquarium (and When Chessify Suffices)

ChessBase 26

ChessBase is the professional standard for database management, opponent preparation, and game annotation. Version 26 adds a revamped opening report that shows win statistics across Elo bands — useful when preparing against opponents at specific levels — along with a reference search that visualises typical piece paths and Monte Carlo analysis showing the most common figurine trajectories from a given position. For correspondence chess the core workflow is: import your opponent’s games from the Corr and Mega databases, run a player-specific filter, study their repertoire with both colours, and build a preparation tree. ChessBase handles this better than anything else available.

Aquarium 2026 (IDEA)

Aquarium 2026 (currently €48.97) earns its place in correspondence chess almost entirely because of one feature: Interactive Deep Analysis, or IDEA. Where ChessBase analysis is essentially linear — you set an engine running and read the output — IDEA builds a branching tree of engine analysis automatically, exploring multiple candidate moves and their replies over extended sessions. You can set it running overnight and return to a comprehensive decision tree rather than a single-line evaluation. For correspondence chess, where you’re often asking “is line A or line B better, and how far does the difference persist?”, this is a meaningful advantage.

The downside: Aquarium is niche, development pace is slower than ChessBase, and the interface feels its age. Many players use ChessBase for preparation and database work, then Aquarium for deep tree analysis on specific positions.

Chessify

Chessify is a cloud analysis platform, not a chess management program, which makes it a different kind of tool. For players who want to access engine analysis from any device, including mobile, or who want to run Lc0 without owning a strong GPU, it fills a real gap. The free tier gives Stockfish at 1 MN/s. The Amateur plan (~$8/month) opens up unlimited shared-server analysis at considerably higher speeds. Lc0 access requires coins or the GM plan. At the top end, coin-based dedicated servers can push Stockfish to 1 billion nodes per second — a level of hardware that would cost a serious sum to replicate locally, available here on demand for cents per hour. And if that still isn’t enough, Chessify has been known to build custom configurations for those with sufficiently specific requirements, though that conversation probably begins with a somewhat larger budget in mind. Chessify has also integrated the ICCF database (2M+ games) at the Master plan level and upward, and can be used as a remote engine source for ChessBase, Fritz, HIARCS, and SCID — integrating with your existing workflow rather than replacing it. For correspondence chess at a serious level, Chessify alone is not a complete solution, but as infrastructure for accessing raw engine power without local hardware constraints, it’s genuinely worth knowing about.

Engines

At the top level of correspondence chess, Stockfish and Lc0 are the two engines that matter. Everything else is either a cross-check or a source of ideas.

Stockfish 18

Stockfish remains the strongest engine by most measures and the workhorse of serious correspondence analysis. Its NNUE evaluation handles everything competently, but it shines particularly in sharp, tactical positions where precise calculation is what the position demands. If a line is being decided by a concrete sequence five or ten moves deep, Stockfish will find it. Run locally on a modern multi-core CPU with adequate RAM it’s fast enough for most correspondence work. For deeper analysis — very long overnight sessions, or critical positions where you need more certainty — Google Cloud is a scalable option for running your own Stockfish instance without being constrained by local hardware.

Lc0 (with network considerations)

Lc0 — Leela Chess Zero — is the neural network engine, and its character is genuinely complementary to Stockfish’s. Where Stockfish is concrete and tactical, Lc0 evaluates positionally in a way that feels more human: it grasps long-term structural factors, piece coordination, and the comparative value of piece configurations in ways that pure alpha-beta search sometimes undersells.

There’s a technical dimension worth understanding here. Lc0 produces not just an evaluation but a policy — a probability distribution over all legal moves representing its intuitive likelihood of each candidate. Moves that are very strong but score a low policy probability are, by that measure, surprising: the kind of move a strong human wouldn’t naturally consider. When Stockfish confirms that a low-policy Lc0 candidate is objectively the best move, you’ve found something close to a genuine brilliancy.

For correspondence chess, though, the most immediately useful divergence is simpler: positions where Stockfish and Lc0 disagree on evaluation rather than on move choice. Those disagreements are flags that demand deeper investigation. Here is a position from my Polish Fischer Random Championship final as an example. After White plays Ng6:

Stockfish 18 evaluated this position at −0.02 — essentially equal, bordering on irrelevant. Lc0 gave −0.85, a significant structural advantage for Black. The position isn’t tactically forcing. What Lc0 is reading, and Stockfish is largely discounting, is something about the long-term relationship between Black’s bishop, the pawn structure, and the activity available to White’s pieces across the next twenty or thirty moves. When two engines of that calibre read a position this differently, the only sensible response is to allocate more time and dig deeper before committing. In a correspondence game you have that luxury; you should use it.

The network you run with Lc0 matters. Larger networks — more parameters, richer positional understanding — are generally preferable for correspondence chess if you have the hardware to run them. Lc0 is GPU-bound, which is its main practical limitation. Vast.ai is the popular cloud solution: GPU instances rentable for cents per hour, allowing you to run Lc0 with a strong network without needing a high-end gaming machine. Chessify also provides Lc0 access via coins or the GM subscription tier.

Other engines

CorChess — a Stockfish derivative specifically oriented toward correspondence chess — has a following in the community and is available on Chessify. In honest practice, it probably doesn’t offer a meaningful edge over Stockfish 18’s main branch, but it’s worth running as a cross-check in critical positions: when CorChess and Stockfish disagree, that disagreement is informative. Similarly, Crystal (which has notable fortress-detection capability in endgames), Caissa, ShashChess, and RubiChess are useful for the same reason. None of them are stronger than Stockfish in general, but they evaluate positions differently, and a position where multiple engines disagree on sign deserves more of your attention than one where everything agrees.

Endgame Tablebases

This is one area where “you should probably know about this” understates the case. Tablebases are part of the infrastructure of the game.

Six and seven-piece: know what you can claim

The Syzygy six-piece set — covering every endgame with up to six pieces including both kings — weighs approximately 149 GB in total and is freely available for download from Lichess servers. On a modern SSD this is a trivial storage investment and a one-time effort. There is no serious reason not to have the full six-piece set installed and configured for your engines. Do it if you haven’t already.

The ICCF formalised the implications: players may claim a win or draw based on Syzygy tablebase evidence — six-piece since 2014, extended to seven-piece following the 2019 Congress. This isn’t merely an analytical tool — it’s a claims mechanism. If your opponent is in a tablebase position that is certified as drawn, you can enforce that result regardless of what the position looks like on the board.

Seven-piece: selective download is the only practical approach

The complete seven-piece Syzygy set totals approximately 16.7 TiB — far beyond what any personal storage setup can reasonably hold. The answer is to be selective, guided by frequency of occurrence.

The top seven-piece endgame by frequency is KRPPvKRP, which arises considerably more often than any other and should always be the first download. The next tier includes KBPPvKBP, KPPPvKPP, KRPPPvKR, and KQPPvKQP. The Komodo/Dragon documentation recommends starting with KRPPvKRP and KPPPvKPP (totalling roughly 35 GB combined), and their Syzygy7.pdf lists the top 48 most common seven-piece endgames — a useful guide for building a selective collection. A practical target for most serious players is the top 20 by frequency, which covers the vast majority of positions you’ll actually encounter.

The download process, done by hand from the Lichess tablebase server, is genuinely time-consuming — there’s no polished tooling that makes it painless, and the files for the top 20 endgames alone run to several hundred gigabytes. A dedicated SSD is necessary rather than optional: engine probing requires the read speeds that spinning disk simply can’t provide. The setup is a commitment. But once it’s done and your engine configuration points at it, the result is silent and perfect: instantaneous knowledge of those positions whenever your analysis reaches them. One of those infrastructure investments that you make once and then forget about in the best possible way.

For material configurations you haven’t downloaded locally, syzygy-tables.info provides free web-based lookup for all positions up to seven pieces — handling the long tail of unusual material that you’ll encounter perhaps once per tournament.

Beyond Tablebases: FinalGen and Alternatives

FinalGen is a dedicated endgame tablebase generator for positions that fall outside Syzygy coverage. Given time and computing resources, it can generate an exact result for a specific position — not an evaluation, not a probability, but a definitive win, draw, or loss. For those needing four-piece depth-to-mate analysis specifically, Hoffman is an alternative that reports DTM (distance to mate) rather than Syzygy’s DTZ metric — useful when you want to know the minimum number of moves to checkmate rather than merely the distance to the next pawn move or capture.

The scenario where these tools matter most in practice isn’t usually “can I win this endgame.” It’s the decision that happens before the endgame: whether to enter it at all. A typical situation: you’re in a complex position and one conversion path leads to a pawn-and-piece endgame that your engines assess as “slightly better,” but you can’t determine whether that edge is real and convertible or whether it collapses into a fortress where your opponent parks their pieces and the win disappears. Running FinalGen on the target endgame position before you commit to the conversion path can give you a definitive answer — which completely changes how you approach the twenty preceding moves.

I’ve had games where exactly this calculation mattered: a moment where simplification was available but the theoretical status of the resulting endgame was genuinely unclear. The engine said “better.” The actual question was “better enough, and can Black hold a fortress?” Knowing the answer before you make the choice is what these tools provide. Like the seven-piece tablebases, you won’t reach for them often. But the positions where they’re relevant tend to be the ones that decide tournaments.

How it Actually Fits Together

The tools described above don’t live in isolation — they form a workflow. Here is roughly how a serious analysis session on a specific position might look in practice.

You start from the position on the board and search for it across your databases — Mega, Corr, OM CORR, and ChessDB.cn simultaneously. You filter the results: recent games only, a minimum rating threshold, no short draws. This gives you the human and engine-validated map of where theory currently stands in that position.

In parallel, you run engine analysis: Stockfish and Lc0 in infinite analysis, engine matches from the position, ChessBase’s Monte Carlo analysis for statistical trajectory, and Deep Position Analysis for slower, broader evaluation. ChessDB.cn serves as an independent cross-check on whatever your local engines are suggesting.

All of this — the filtered game sets and the engine lines — feeds into Aquarium’s IDeA project. IDeA seeds itself with these inputs, adds any new positions it encounters as tasks, and runs overnight (or for days, in critical positions), continuously updating the branching evaluation tree. You return to explore the results, identify the lines that demand further attention, expand them, and repeat.

The loop is: search → analyse → seed IDeA → explore → repeat, with ChessDB.cn as a sanity check throughout. It sounds systematic because it is. Correspondence chess at a serious level is closer to research than to game-playing, and this is what the research process looks like.

Putting it Together

A sensible configuration for a serious correspondence player today: core theoretical books with their key claims verified against current engines; Mega and Corr from ChessBase plus OM CORR for correspondence coverage and ChessDB for free engine-validated lookup; ChessBase 26 for preparation and database management, with Aquarium for IDEA tree-building; Stockfish locally as the workhorse, Lc0 via Vast.ai or Chessify for positional cross-checking, other engines as disagreement detectors; six-piece Syzygy in full, seven-piece downloaded selectively by frequency with syzygy-tables.info for the rest; and FinalGen for the position you can’t resolve any other way.

The total investment — in money, in setup time, in the learning curve — is real. But then, so is the game.

Amici sumus — Paweł

PS: All tools, databases, and services mentioned in this article are ones I use or have evaluated myself. I have no affiliate relationships with any of them, receive no commission or compensation for links, and paid for everything out of my own pocket — or found it free. The prices quoted were accurate at time of writing but may change.

Sean Clark wrote this piece for the EFCC Magazine, which was unfortunately discontinued before it could be published. I am glad it has found a home here instead. It is aimed at those already familiar with chess who are new to correspondence play — and it addresses, from a personal rather than technical perspective, the question every correspondence player must eventually answer for themselves.

— Paweł

The answer to this question is a journey, and all journeys have a beginning if you will indulge me.

I first encountered chess as an eight-year-old at my uncle’s house, learning the rules under his guidance. Growing up in the 1970s and 1980s, I witnessed the chess world undergo some dramatic changes. Bobby Fischer was fading into obscurity, Garry Kasparov was rising to fame, and Nigel Short was creating a stir as Britain’s prodigious young talent. Chess even made it to mainstream media with The Master Game, a BBC series that sought to popularise this ancient art.

Despite the excitement around the game, I grew up in a small village where chess wasn’t exactly a part of daily life. My world was filled with fantasy novels, Dungeons & Dragons, and sci-fi escapades. Living in a country pub, I found ways to entertain myself indoors in winter and outdoors in the warmer months. Chess was a fleeting interest, with no one nearby to play. Chess clubs were out of reach, there was no internet, and my school lacked a chess program. Besides, my family wasn’t exactly brimming with chess enthusiasts.

That all changed when I acquired my first computer in the 1980s: the glorious ZX81 with its impressive 16KB of expanded memory. Suddenly, I had an opponent at my fingertips — although a silicon one, it was at least willing. With no tutor or exposure to chess literature, my learning was piecemeal at best, but I enjoyed the occasional match against the computer.

As I moved into my teens, golf took centre stage, and chess receded. And then came the usual suspects: studying, working, and, of course, girls! Chess became a puzzle, something to dabble with here and there, but nothing I took particularly seriously. I was still drawn to physical chess sets, buying the occasional cheap wooden set from holiday spots like Greece and Spain. However, these beautiful boards often ended up gathering dust due to a perpetual shortage of opponents — a common fate for many holiday chess purchases, I suspect.

The internet, however, changed everything. With the advent of online platforms, I could play any time against opponents worldwide, and the anonymity suited me. In traditional chess clubs, I felt out of place. Members were either gifted youngsters or seasoned veterans who had been at it for decades. I was neither, and while I loved the game, deep analysis and memorised positions were beyond me. Online, though, I could just enjoy the game, free from in-depth discussions on positional play that, frankly, went over my head.

Time was my other nemesis. Bullet games — those exhilarating one-minute blitz rounds — were a thrill, but I lost time and again due to my lack of instinctual play. Correspondence games, with their multi-day time controls, were a revelation. Here was a game I could play at my own pace. However, even there, I found some opponents abandoning their games halfway, which made the victories feel hollow.

Then I discovered ICCF and EFCC, the formal world of correspondence chess. Now, this was serious. Titles were on offer, and I could represent my country in international competitions. My first invitation to play for England filled me with pride. I told everyone I knew. My uncle, the very man who taught me to play, was especially thrilled. After all, he had taught the first person in our family to represent their country in any sport at any level! It was a surreal moment, as though a pastime I had cherished privately was suddenly validated.

In correspondence chess, I could delve into the richness of each position, take my time, and compensate for my lack of chess knowledge with computer assistance. I used engines, not as a crutch, but as a guide through complex positions. What felt like cheating at first soon became a valuable tool in my arsenal. And, as I learned the hard way, blindly following engine moves without understanding why doesn’t guarantee victory.

In recent months, my pursuit has become more serious as I work toward my first title. I approach each game with a much deeper analysis, going beyond superficial moves to uncover strategic depths. I prefer aggressive lines if the odds favour me, even if it’s not the engine’s “first choice.” With time on my side, I stick to a select few openings, diving into intricate positions and developing a true appreciation for the nuances within.

Engines no longer feel like an affront to the purity of the game; instead, they are an enhancement, helping me spot patterns I missed in my younger days. Today, I no longer wonder, “What’s the right way to use an engine in correspondence chess?” because I’ve learned that each player must find their own answer. Engines, much like chess itself, can be as rigorous or relaxed as you choose to make them.

My approach to engine use varies by game. I often use conditional moves to clarify a draw or signal a well-prepared sequence to my opponent. Why continue a game that’s all but decided when there are countless others to savour? It may be a bit cheeky, but it’s also the freedom I missed in my early years with the game.

One final note on engines: everyone playing on ICCF can access them. Despite this, draws are common, and there are still those who lose despite the advantage of technology.

At the end of the day, there is no right way to use an engine in correspondence chess. It’s about enjoying the game your way.

— Sean Clark

Further reading

While on the subject of correspondence chess resources: I also want to draw attention to the blog of Christer Sundqvist, an International Correspondence Chess Master from Helsinki, at chessturpaduunari.blogspot.com. Christer has produced extensive analysis packages on the f3-Nimzo-Indian, the Sicilian Najdorf, and the Queen’s Gambit Accepted — all developed through the lens of correspondence chess, where theory genuinely goes deeper than anywhere else. If you play any of those openings and want to see what serious correspondence preparation looks like, his blog is essential reading. It is exactly the kind of resource The Sealed Move exists to surface.

There is a particular kind of tension unique to correspondence chess finals. It does not arrive in a single moment — no time scramble, no blundering under pressure, no dramatic clock flag. It accumulates over months, sometimes years, in the quiet space between moves. And occasionally, after all that time, it resolves in a way that feels almost arbitrary.

The 3rd Polish Chess960 Championship Final (POL/C960-3/f) was one of those tournaments.

Ten players. Forty-five games. A start date of February 2025 and an end date officially set for June 2026. Each game played under the Triple Block system with 50 days initial time bank and 3 days increment per move — enough time, in theory, to find the truth in any position. In practice, enough time to find the truth and still not separate five players who finished level on every meaningful measure.

When the dust settled, five of us had scored 5 points from 9 games. One win each. Sonneborn-Berger: 21 points apiece. Direct encounters: inconclusive. The tournament rules, written in Polish and English on the ICCF server, were unambiguous about the final tiebreaker: wyższy ranking na koniec turnieju — higher rating at the end of the tournament on the ICCF Fischer Random list.

I, rated 2563, took first. Adam Sazon, with 2526, finished second. Three players behind us — Jerzy Staniszewski, Bronisław Bartkiewicz, and Wojciech Krzyżanowski — completed the five-way tie, separated from each other by the same merciless decimal places.

This is the reality of high-level correspondence chess in the mid-2020s. In an era of engine-assisted play, where both players have access to the same analytical tools, decisive results have become increasingly rare. The draws accumulate. The tiebreakers sharpen. A championship can be decided not by the quality of your chess over eighteen months but by a rating figure calculated from results elsewhere.

It is, in a word, unsatisfying — and it is worth being honest about that.

And yet.

The one game I won in this tournament was not a gift. It was not a blunder by my opponent, not a time forfeit (well, technically it was, keep reading), not a fortunate oversight in a balanced position. It was a long, difficult, technically demanding game that required navigating genuine complexity over many moves and many weeks. The win did not feel like luck at the time. Looking back at it now, it still does not.

That game is what this piece is about.

The game

My opponent was Jacek Klimecki, rated 2358 — the lowest-rated player in the field, but as the final standings showed, ratings in correspondence chess tell only part of the story. Every game in this tournament was played with full engine assistance. Rating differences narrow when both players have access to the same analytical tools.

The starting position was Chess960 #761. What matters for our purposes is that neither player had any opening theory to lean on. In Chess960, the pieces are randomised behind the pawns, and with randomisation comes a particular kind of equality: both players must think from move one.

The opening: engines uncertain, humans navigating

The first eleven moves were, by correspondence chess standards, relatively conventional — both sides developing sensibly, staking out central space, probing for weaknesses without committing to anything irreversible.

1.e4 a5 2.a4 d6 3.Ne3 e6 4.h4 Nb6 5.c3 h5 6.g3 g6 7.f4 f5 8.Bf3 c6 9.d4 e5 10.dxe5 dxe5 11.Nd3 Nfd7

This is one of the genuinely interesting features of correspondence Chess960 at the highest level. After eleven moves, Stockfish 18 — one of the strongest engines in existence — evaluates the position at exactly 0.00. A draw, somewhere in there. The engine sees several reasonable continuations for White: 12.Bd1, 12.Bf2, maybe even 12.exf5 all maintain equality. The position is balanced, complex, and theoretically unexplored.

What my opponent played was 12.Kc2.

Move 12: where the hunt begins

12.Kc2 is probably still a draw with precise play. Stockfish 18 evaluates it at approximately -0.55 — a solid positional advantage for Black, but nowhere near winning. No tactics. No blunders. Just a slightly uncomfortable structure for White that will require accurate defence over many moves to hold.

In over-the-board chess, -0.55 is often nothing. Players at all levels hold positions with that kind of deficit every day. The clock runs, opportunities pass, the game simplifies to a draw.

In correspondence chess, -0.55 is an invitation.

With fifty days per move and engine assistance on both sides, a persistent structural advantage does not evaporate. It must be defended, move by move, over weeks and months, against an opponent who has all the time in the world to find the best continuation. The question is not whether the advantage exists — the engine confirms it does — but whether it can be converted. That is still a human question. And it is where the game became interesting.

The hunt begins.

Move 15: the second mistake, and what -1.00 actually means

Play continued 12...Bf7 13.Bf2 fxe4 14.Bxe4 Bc7.

White has a decision to make. The engine is unambiguous about what is required: 15.Nf1, heading for d2 via a regrouping manoeuvre — slow, patient, defensive. The knight on f1 eyes d2, stabilising the position and preparing to challenge Black’s growing initiative. It is not an exciting move. It is a necessary one.

The evaluation after 15.Nf1 remains approximately -0.55. The position is still holdable. The task is difficult but not impossible.

My opponent played 15.b3.

Stockfish 18 evaluates the position after 15.b3 at -1.00.

This number requires explanation, because in correspondence chess it means something specific and significant.

In the standardised Stockfish scoring system, -1.00 represents the point at which the engine would win approximately half of its games against another instance of itself from that position. Not all of them — the position is not lost by force, there is no immediate combination, no tactical refutation. But the structural and dynamic factors have accumulated to the point where, with best play on both sides, Black converts more often than not even against perfect opposition.

In over-the-board chess, two players at human speed might well draw from -1.00 many times. Fatigue, time pressure, and the sheer difficulty of finding the best continuation across dozens of moves all work in the defender’s favour.

In correspondence chess, they do not. With fifty days per move and an engine running on both sides, -1.00 is close to a verdict. Not certain — the game is not won in any obvious way, and a great deal of manoeuvring remains. But the margin for error has narrowed to a point where White must play with near-perfection for many moves in a row, while I need only maintain the pressure and wait.

Two significant inaccuracies in a correspondence game — 12.Kc2 and now 15.b3 — is too much to hope to escape from. The question is no longer whether Black has an advantage. The question is how to convert it.

That process would take another fifty moves.

Move 30: where engines go quiet and humans must decide

The middlegame that followed was exactly what -1.00 promises and nothing more — a long, grinding process of tightening the position, preventing counterplay, and steering towards an endgame where Black’s structural advantages could be expressed. Neither dramatic nor obvious. The kind of chess that correspondence play is built for.

After 30.Kxc3, the engine fell silent in a different way.

Two moves present themselves for Black. 30...Qc5+ and 30...Qd4+. Stockfish 18 evaluates both at approximately the same figure — the distinction between them, in terms of raw centipawns, is negligible. The engine sees equivalent winning chances in either direction and declines to express a strong preference.

This is the moment in correspondence chess that no engine can resolve for you.

Both moves lead, with best play from both sides, to an endgame featuring two bishops against two knights. The material is roughly balanced. The question is which version of that endgame — which pawn structure, which king position, which specific configuration of pieces — gives Black the better practical chances. Not better in theory, necessarily. Better in practice, against this opponent, from this position, in this tournament.

I spent a considerable amount of time here. Not calculating — the engine handles calculation. Thinking. Going deep into the lines, tracing where my opponent might find rescue, where he might face a harder choice, where a mistake was more likely to appear. Both paths seemed to lead to the same general destination, but the roads themselves were different.

In the end I played 30...Qc5+.

I will not pretend it was a purely analytical decision. The lines were close enough that something else entered — call it intuition, or pattern recognition, or simply a judgment about which type of endgame I understood better. 30...Qd4+ was equally valid. I chose Qc5+ because the positions it produced felt, to me, more concrete. More navigable. More likely to offer a moment of clarity in the endgame to come.

Is two bishops against two knights winning in theory? The bishops are more mobile across an open board — that much is well established. But who can know for sure in this specific configuration, with these specific pawns, at this specific stage of the game. The engine offers a number. The conversion remains a human problem.

What followed was fifty more moves of technical endgame chess — the kind that correspondence play was invented for, and the kind that almost no other format of the game could produce.

The endgame: a dark forest with one way out

After 36...Bxd8 the queens and rooks were gone. What remained was the endgame I had been steering towards since move 30 — two bishops against two knights, with pawns on both sides of the board.

This is where correspondence chess earns its reputation for brutality. Not the tactical brutality of combinations and sacrifices, but the slow, methodical brutality of technique. An endgame that a computer evaluates as winning in principle must still be converted move by move, over weeks, against an opponent who has the same engine and the same amount of time.

My approach had two components.

The first was practical: I was watching the clock. In correspondence chess, time pressure is a slow accumulation — not seconds but days, draining from a bank that started at fifty and does not refill quickly. Keeping the position complex, forcing my opponent to spend time on difficult decisions, was itself a weapon. A player running low on time makes worse decisions. The pressure is quieter than a ticking clock but no less real.

The second was the dark forest principle. There is a famous observation in chess — attributed, in various forms, to several grandmasters — that the best way to win a game is to lead your opponent into a dark forest where there is only enough of a path for one. In endgame play, this means creating positions of sufficient complexity that the engine’s recommended defence is narrow, non-obvious, and easy to miss under pressure. Not positions where the opponent blunders a piece. Positions where the correct move looks wrong, and the wrong move looks natural.

There were several of those traps in the endgame that followed. Most were escaped — my opponent defended well, and I do not want to suggest the win was straightforward. It was not.

The final trap arrived after 59...Ba7+.

White has two king moves. 60.Kc4 is the only way to hold — a counter-intuitive retreat that keeps the king active and prevents Black’s king from entering the kingside. 60.Kd3 looks natural, centralising the king, but it allows my king to infiltrate via the kingside pawns, and from there the position collapses.

My opponent played 60.Kd3.

The rest, as they say, is history. My king advanced, the pawns fell, and the position became objectively lost. The game was eventually decided on time — my opponent’s clock ran out before the position reached its logical conclusion — but there was no rescue available regardless. The forest had only one path out, and White had taken the wrong one.

A note on opponents and what comes next

I want to be clear about something before closing.

Jacek Klimecki played well. All of my opponents in this tournament played well. The five-way tie on 5 points, decided by rating in the final tiebreaker, reflects the extraordinary quality of the field and the difficulty of the format. In Chess960 correspondence chess at the national championship level, decisive results are earned, not given.

The championship itself was a prelude. This year, the 4th Polish Chess960 Championship will be held — an international tournament, open to players beyond Poland, organised to commemorate the 100th anniversary of the Polish Chess Federation. It is a significant event, and one that attracted the strongest correspondence Chess960 players from across the ICCF world.

To my opponents in the 3rd championship, and to those who will compete in the 4th: amici sumus.

We are friends. The game is the thing.

Full game score:

Chess960 Position #761
White: Klimecki, Jacek (2358) — Black: Fiedor, Paweł (2437)
3rd Polish Chess960 Championship Final (POL/C960-3/f)
Result: 0-1 (65 moves)

1.e4 a5 2.a4 d6 3.Ne3 e6 4.h4 Nb6 5.c3 h5 6.g3 g6 7.f4 f5 8.Bf3 c6 9.d4 e5 10.dxe5 dxe5 11.Nd3 Nfd7 12.Kc2 Bf7 13.Bf2 fxe4 14.Bxe4 Bc7 15.b3 O-O-O 16.Rge1 Rge8 17.Rad1 Kb8 18.Qg2 Qf8 19.Kb2 Qe7 20.f5 gxf5 21.Nxf5 Qe6 22.Nc1 Rg8 23.Re2 Nf6 24.Rxd8+ Rxd8 25.Nh6 Be8 26.Bf5 Qd6 27.Be3 Nbd5
28.Bg5 e4 29.Bxe4 Nxc3 30.Kxc3 Qc5+ 31.Kb2 Qd4+ 32.Kc2 Nxe4 33.Qxe4 Qxe4+ 34.Rxe4 Bg6 35.Bxd8 Bxe4+ 36.Nd3 Bxd8 37.Kd2 Bd5 38.Kc2 Bf6 39.Nf4 Bf3 40.Nf7 b5 41.Kd2 b4 42.Ng5 Bg4 43.Ne4 Bd4 44.Kd3 Be5 45.Ng6 Bc7 46.Nf4 Kc8 47.Kd4 Bb6+ 48.Kd3 Bd1 49.Kc4 Bd8 50.Nf2 Bc2 51.N2d3 Bd1 52.Nf2 Bf3 53.Kd3 Bb6 54.Ne4 Bd1 55.Kc4 Bc7 56.Nd2 Bb8 57.Ne4 Bf3 58.Kd3 Bg4 59.Kd4 Ba7+ 60.Kd3 Bd1 61.Kc4 Kd7 62.Nf6+ Ke7 63.N6xh5 Kf7 64.g4 Bxg4 65.Ng3 Bf2 0-1

Engine evaluations thanks to Chessify.

There is a move sitting on a server in Germany right now that has been waiting three weeks for a reply.

The player considering it has a position open on his screen. He has run it through an engine, consulted databases of previous games, slept on it, changed his mind twice, and is now reconsidering a variation he dismissed a fortnight ago. He is not in a hurry. His opponent, somewhere in another country, is doing the same thing with their own positions in their own time. They may never meet. They are, in a meaningful sense, playing the same game of chess — just not at the same time.

This is correspondence chess. And if you have never heard of it, you are not alone.

A game without a clock

Most people who know chess know it as a timed game. You have minutes, sometimes seconds, to find your move. The clock ticks. You decide. The pressure is part of the drama.

Correspondence chess removes the clock entirely — or rather, replaces it with something almost unrecognisably different. Players are typically given between 10 and 50 days per move. Some tournaments allow even more. A single game can last years. A world championship match conducted by correspondence has historically taken the better part of a decade to complete.

The effect on the chess itself is profound. Decisions that in over-the-board play must be made in minutes — under pressure, imperfectly, with incomplete calculation — can in correspondence chess be examined to their logical conclusion. Every variation can be followed to its end. Every assumption can be tested. The result, at the highest levels, is chess of a depth and precision that no over-the-board game can match.

How it began

Correspondence chess is older than the telegraph. The first recorded correspondence game dates to 1804, played between two Parisian chess clubs by exchanged letters. By the mid-19th century, matches between cities and countries were commonplace — London versus Edinburgh, England versus France. Each move was written on a postcard or letter, posted, and awaited.

The pace was not a limitation. It was the point. Players had days or weeks to think because the post took days or weeks to arrive. The game adapted to its medium.

As communication accelerated, so did correspondence chess. Telegrams replaced letters for important matches. Email eventually replaced telegrams. And then, in the 1990s and 2000s, dedicated servers emerged — platforms where players could submit moves, track games, and compete internationally without ever exchanging a physical letter.

Today the International Correspondence Chess Federation, the ICCF, administers tens of thousands of games simultaneously across its web server. Players from over 60 countries compete in national team championships, thematic tournaments, and individual title events. The World Championship is a multi-year, multi-stage competition contested at a level of theoretical preparation that would be unrecognisable to the players who posted their moves by hand two centuries ago.

The postcard is gone. The patience remains.

The engine question

The arrival of strong chess engines — software that can calculate millions of positions per second and plays well beyond any human — changed correspondence chess irrevocably. The ICCF permits engine use in most of its events. This is not a concession to cheating; it is an acknowledgement of reality, and a philosophical choice about what correspondence chess is.

The argument for permitting engines runs like this: correspondence chess has always been about finding the best move with the resources available. In the 19th century, those resources were books, analysis boards, and time. Today, they include engines. The challenge has not disappeared — it has shifted. With engines, both players have access to the same raw calculation. What distinguishes them is preparation, understanding, and the ability to navigate positions where engines themselves are uncertain.

In practice, the highest levels of correspondence chess have become a dialogue between human understanding and machine calculation. The engine tells you what the computer thinks is best. The correspondence player must understand why, must navigate the transition from opening to middlegame to endgame, and must make judgments in positions where the engine’s evaluation wavers — where the silicon is, in a word, unsure.

Those are still human decisions. They are simply better-informed ones.

What correspondence chess produces

The games produced at the top of correspondence chess are unlike anything else in the sport. Theoretical novelties — new ideas in known opening positions — appear in correspondence games years before they reach over-the-board play. Endgames are conducted with a precision that approaches the theoretical ideal. Middlegame combinations are calculated to depths that would be impossible for any human mind working under time pressure.

The ICCF database contains millions of correspondence games. It is, in a real sense, the deepest record of what chess actually looks like when humans and machines work together without constraint. Grandmasters preparing for over-the-board tournaments study correspondence games for precisely this reason: they show what the position truly demands, without the noise of time pressure and fatigue.

Some of the most beautiful games in chess history were played by correspondence. They are almost entirely unknown to the general public. That is one of the things this publication exists to change.

Who plays it

Correspondence chess has always attracted a particular kind of player. Not necessarily the strongest over-the-board player — though many strong players have competed by correspondence — but the most analytical, the most patient, the most interested in chess as an intellectual exercise rather than a competitive spectacle.

It has also, historically, been genuinely international in a way that over-the-board chess was not. During the Cold War, correspondence chess was one of the few arenas where players from the East and West competed on equal terms, their moves crossing borders that their bodies could not. National team championships — the ICCF Olympiad — were bitterly contested between the Soviet Union and the West for decades. The games themselves were diplomatic dispatches of a kind.

Today the game is smaller than it was at its peak, but the community is serious, dedicated, and largely invisible to mainstream chess culture. There are no correspondence chess streamers. There are no correspondence chess influencers. There is, as far as I am aware, no publication devoted to it for a general audience.

Until now.

A personal note

I have been playing correspondence chess for some time now. I hold the title of Senior International Master at the ICCF — one of the federation’s senior competitive titles — and I am the current Polish national champion in Fischer Random chess, the variant in which the starting position of the pieces is randomised, eliminating opening preparation and testing pure chess understanding.

I am also Polish, which brings me to a small digression worth making.

This year marks the 100th anniversary of the Polish Chess Federation — a century of organised chess in a country with a proud and sometimes dramatic chess history. The federation has begun its anniversary year with a commemorative tournament, details of which are here. For anyone with an interest in Polish chess, or in the broader history of European chess organisation, it is worth reading.

Correspondence chess rewards patience. So does this publication. Posts will arrive when there is something worth saying.

If you play correspondence chess, have played it, or are simply curious about what serious chess looks like when time is not a constraint — you are in the right place.

Paweł Fiedor — Senior International Master, ICCF & Polish National Champion, Fischer Random Chess