Every number, accounted for.

What we compute

The engine behind all four tools is the pair of tables FIDE publishes with its Rating Regulations. Table 8.1.2 maps a rating difference to an expected score in 0.01 steps - D = 0 gives 0.50, D = 200 gives 0.76, D = 400 gives 0.92, and differences above 735 give 1.00. Table 8.1.1 is its mirror: a percentage score to a rating difference, from 0 at 50% up to 800 at 100%. We read both tables literally - no interpolation, no smoothing.

Where a continuous value is useful (the curve on the expected-score page, for instance) we also compute the exact logistic function E = 1/(1 + 10^{(R_opp−R)/400}) that underlies the whole Elo family[W1]. The two never differ by more than rounding: at D = 400 the table says 0.92, the curve 0.909.

Rules we implement

• K-factor (art. 8.3.3). K = 40 for a player new to the list until events with at least 30 rated games; K = 40 until the end of the year of the 18th birthday while rated under 2300; K = 20 below 2400; K = 10 once 2400 is published - permanently. We also honour the K × games ≤ 700 cap in the auto-K explanation.
• 400-point rule (art. 8.3.1). For players rated below 2650, a difference beyond 400 counts as exactly 400 (as amended from 1 October 2025); 2650+ ratings use the true difference.
• Rating change (art. 8.3.2–8.3.4). Per game, ΔR = K × (S − E). For a rating period, FIDE sums (S − E) over the period's games, multiplies by K, and rounds to the nearest whole number with 0.5 away from zero. Our "official-style" totals follow this; sequential projections are labelled as such.
• Performance rating. TPR = Ra + dp (FIDE method, art. 8.2 uses the same machinery for initial ratings, including the two hypothetical 1800 draws and the 2200 cap); the alternative "rating + score-based delta" anchors the same dp to your own rating; the linear "algorithm of 400" is shown for reference[W1].
• Norms context (Title Regulations art. 1.4). GM norm: 2600+ performance over 9+ games vs opponents averaging ≥ 2380; IM: 2450+ vs ≥ 2230; with the titled-opponent and federation mix as stated[F2]. We show these as benchmarks - awarding norms is FIDE's job, not ours.

What we deliberately do not do

We do not estimate win/draw/loss splits (Elo pins the expected total, not the shape), we do not fetch or store any personal data - every calculation runs locally in your browser - and we do not pretend a simulation is an official rating: FIDE settles ratings once per period, by list, with the arbiters' reports in hand.