Auditing your own rating: the four leaks holding your Elo back

Short answer

A rating audit is a statistical reading of your own games to find where the points go — not another study plan. Four leaks account for most plateaus: training in one time control and competing in another, losing games in the conversion phase, concentrating your losses in a handful of opening lines, and playing on after the first loss of a session. This guide shows how to measure each one and what to do with the answer.

The essentials in a few lines

• Auditing is measuring, not studying: the question is not 'what don't I know', it is 'where are the points I win being handed back'.
• Leak 1 — the wrong sample: training in bullet and competing in classical means measuring one skill while improving another.
• Leak 2 — the conversion phase: most points lost by players between 1200 and 1900 come out of positions that were already winning.
• Leak 3 — the concentrated repertoire: a minority of lines usually accounts for the majority of losses, and it is the cheapest list to fix.
• Leak 4 — the long session: your score drops game by game after the first loss, and this is the leak you close without studying anything.

What auditing your rating means

The difference between studying and auditing shows up in the opening question. Studying answers 'what do I still not know', a list that never ends and that any player can fill in five minutes. Auditing answers 'where exactly are my points going', a short, specific list with opening names and move ranges — and it usually reveals that the topic you were about to study was not among the causes.

Any audit rests on a sufficient, homogeneous sample: your last hundred games in the time control you actually want to improve. Fewer than that and noise dominates; mix time controls and you measure an average that corresponds to none of the skills. It is the same requirement as any measurement — and the reason most conclusions players draw from their own games are wrong.

It is worth saying what an audit does not do. It does not tell you whether you are good or bad, it does not replace training, and it produces no conclusions about talent. It produces four numbers and a list of lines, and its value is in pointing the study hours you were going to spend anyway at the place where they return points.

Leak 1: you train in one time control and compete in another

The measurement is simple and usually uncomfortable: add up your games of the last three months by time control and work out the proportion. When the control you want to improve accounts for less than a quarter of the volume, the plateau is no mystery — most of your chess time is going into an adjacent skill, and the transfer between them is real but partial.

This does not make blitz useless. It is excellent for running a repertoire, exposing opening holes quickly and keeping the habit alive in busy weeks. The problem is proportion and function: blitz as a diagnostic tool pays off; blitz as the main training of someone competing in classical produces exactly the profile of player who moves too fast in positions that deserved fifteen minutes.

The fix does not require a revolution. Swapping two blitz sessions a week for one long game with analysis afterwards already shifts the proportion enough to show up in the next quarter's measurement — and it is the intervention with the best effort-to-return ratio on the whole list.

Leak 2: the phase where the game was already won

To measure it, filter the games where you had a clear advantage at some point and look at how they ended. The number that matters is not how many you won, but how many whole points were handed back after the peak of the advantage — and in which range of moves it happened. In most amateur histories the concentration is obvious and falls between move twenty and move thirty-five.

The causes are usually three, in order of frequency. First, the clock: reaching a winning position short of time and misplaying the technical part. Second, no simplification plan: with extra material, trading pieces is almost always right, and it is the habit least present in players who learned chess by attacking. Third, endgame technique itself, which a handful of well-known basic positions resolves.

The most direct fix is to train the end of the game rather than the start. One weekly session playing only already-winning positions against the engine, with a clock, addresses all three causes at once — and it is the training that plateaued players have typically never done even once.

• Measure your conversion rate: of the games with a clear advantage, how many became full points.
• Note the move range where the advantage evaporates; it tends to be narrow and constant.
• With extra material, trade pieces — the default plan that costs the most points when ignored.
• Save clock for the technical phase; a material edge does not compensate for a lack of time.
• Train basic endgames to automaticity: they are the floor of conversion.

Leak 3: a few lines account for most of the losses

Concentration is the rule rather than the exception: in most histories, a minority of lines accounts for a disproportionate share of the losses. That is good news, because it means the fix is small and local — usually one or two variations with a misunderstood middlegame plan, not the whole repertoire.

When you look at the problem line, resist the automatic conclusion that theory is missing. Check first at which move the evaluation drops: if it drops at move eight, it really is the opening; if it drops at move eighteen, what is missing is the plan for the structure, and studying more variations will spend hours without moving the number. That distinction decides the right kind of study and is the part of the audit that saves the most time.

There is also the repertoire decision proper. If a line has been costing points for months, produces positions you dislike playing and resists two rounds of correction, switching is legitimate — but one line at a time, and with at least thirty games in the new one before judging. Replacing the whole repertoire at every plateau is itself one of the causes of plateaus.

Leak 4: the session that continues after the first loss

To measure it, sort your games by session and compare your score in the first game of each session with your score in the games played right after the first loss. The gap is usually large enough to need no sophisticated statistics, and it translates directly into rating points — because those are games played against opponents in your band with a fraction of your real ability.

The mechanism is well known and nothing specific to chess: after a loss, players tend to move faster, accept complications they would not accept in another state, and chase the lost points back immediately. It is the same pattern that appears in any competitive activity with fast repetition, and online play, where the next game starts in ten seconds, is the perfect environment for it.

The rule that works is simple and mechanical, because decisions taken in that state are not trustworthy: decide before the session how many games you will play, and stop after two losses in a row regardless of how you feel. If you want one more step, analyse the loss before playing again — five minutes with the engine usually removes the urge for an immediate rematch.

• Decide the number of games before the session starts, not during it.
• Stop after two consecutive losses — a rule, not a judgement call in the moment.
• Compare your score in the first game of a session with the last ones.
• A loss on time or to a blunder deserves a longer break than a well-played loss.
• If you do play again, analyse first; the review removes much of the rematch impulse.

Online rating, FIDE rating, and what to compare with what

FIDE's official lists are published periodically and recalculate ratings from games in rated tournaments over the period, with an adjustment coefficient that varies with a player's experience and rating band — players with few games swing more, players in high bands swing less. The exact details live in FIDE's own regulations, and they are worth consulting before drawing conclusions about a specific change.

The practical consequence for anyone auditing is about patience: with a small number of rated games per year, a swing of a few dozen points between lists is within the expected noise. Drawing conclusions about your method from two consecutive lists is reading noise; the reasonable horizon for judging a change in training is a semester's worth of games.

The comparison that genuinely pays is qualitative. Analyse your over-the-board and online games with the same method and split the errors into two groups: those that appear in both environments are chess errors and call for study; those that appear only over the board are process errors — clock, stamina, tension — and call for tournament preparation. They are different diagnoses, and confusing them is the most common way to study the right thing for the wrong problem.

Frequently asked questions