Time Control and Playing Strength in St. Louis
Building on recent results showing the relationship between per-move XPA and Elo ratings, I wanted to take advantage of the recent tournaments in St. Louis to answer the question: how much playing strength do top players actually lose when they play fast?
XPA suggests the answer is about 450 points weaker, in classical terms, when playing blitz chess. This is true not only for the recent events in St. Louis, but also in a broader view across my corpus of analyzed games.
Eight players, three clocks
The recent events in St. Louis provided an interesting matched sample across three time controls. There were rapid, blitz, and classical events, and eight players participated in all of them; not quite the entire field, but close! These events were close in time, with highly similar opposition. We can observe the XPA differences between them, and use the relationship between XPA and Elo (the grey line, based on 94,429 analyzed games with Elo) to translate these differences into Elo language.
Putting these three events on the XPA / Elo curve, we can see:
- The classical event (Sinquefield Cup) has average XPA corresponding to about 2800 on this curve! Note that this curve is built from all games in the dataset, so this higher-than-Elo performance is consistent with rating deflation.
- The rapid event (Saint Louis Rapid & Blitz 2026: Rapid) lands about 150 points lower, near 2650 in classical terms.
- The blitz event (Saint Louis Rapid & Blitz 2026: Blitz) drops considerably more. With average XPA per move dipping to −0.41, it looks statistically like a 2350 classical performance.
Of course, it’s not really as simple as this. We’d expect these top players to retain many of their advantages in faster settings, like their finer positional understanding. And it likely affects different players in different ways. In the following figure we can test that a bit with the data from St. Louis. The overall average is shown along with lines for each of the eight players common to the three events (hover over to see who).
A broader archive check
So far there aren’t a ton of non-classical games in the database. And the sample I do have may be biased, for example corresponding to short tie-break matches. This can be fixed in the future, but it’s still worth asking if the results in St. Louis are representative of this broader archive. To assess this I looked at every analyzed game since 2015 by players rated over 2600. On this wider set of 22,330 games, the gap from classical to blitz across the full set matches the St. Louis gap very closely. A lot more to do here! But this is a nice early signal that the result is reasonable.
| St. Louis 2026 | Archive, 2600+, since 2015 | |||
|---|---|---|---|---|
| Time control | XPA / move | Δ vs classical | XPA / move | Δ vs classical |
| Classical | +0.210 | — | +0.122 | — |
| Rapid | +0.020 | −0.190 | −0.197 | −0.319 |
| Blitz | −0.408 | −0.618 | −0.490 | −0.612 |