From Lichess to the Board: Converting Online Blitz Ratings to FIDE Ratings

1. Introduction

Where would I stand in an OTB tournament with my Lichess rating?

This is a question many players ask themselves. In any competitive sport, knowing how one compares to other participants is of great interest. Sometimes it is even the main purpose, as it determines rankings and prize money.
There have been discussions [1] and previous studies [2][3] attempting to estimate a FIDE rating based on a player’s Lichess strength. I propose an additional method, using data from Lichess profiles where users state their FIDE rating. Based on this data, I fitted a formula to estimate a player’s FIDE rating from their Blitz rating on Lichess.
In the remainder of this paper, I describe the data collection, the analysis and results, and compare the findings with previous work. I also discuss limitations of the approach.

If you are only interested in estimating your own FIDE rating, you may skip ahead to Section 3.

2. Data collection & Analysis

Lichess allows users to share a range of chess-related information in their profiles. One field is reserved for entering a FIDE rating. Using the Lichess API, I downloaded 16,821 active profiles, focusing primarily on players with a Blitz rating above 2,000. Of these, 16.3% included a self-reported FIDE rating.
For the analysis, I applied several filters. Profiles had to show a Blitz rating of at least 2,000, a rating deviation below 100 (to ensure the rating was current), and an absolute difference between Lichess Blitz rating and FIDE rating of less than 450 points (to reduce outliers). I discuss this filtering further in the limitations section.
After filtering, the dataset contained 1,606 profiles. The relationship between reported FIDE ratings and Blitz ratings was analysed using linear regression.

3. Result and Formula

The following formula was estimated:
Est. FIDE Rating = 220.64 + 0.822 * (Your Blitz Rating)
Thus, a player’s FIDE rating can be estimated by multiplying their current Lichess Blitz rating by 0.822 and adding 220.64.
Note, however, that the formula was fitted using players with Blitz ratings above 2,000. For players below that threshold, the results are less reliable.
For those interested in the statistics: F(1,1604) = 1,740.00, p < .001, with an adjusted R2 of .52. Both the predictor and the intercept are highly significant (p < 0.001).

4. Comparison

To compare the results with ChessGoals [3], I aligned my data with Table 3 of their analysis. A comparison to the Universal Rating Converter [2] was not possible, as that method is based on the Lichess Classical rating. The comparison was limited to Blitz ratings above 2,000.

The first and second columns reproduce part of the results from ChessGoals, while the third column shows the FIDE rating estimated using the formula presented here, calculated from the first column. The fourth column reports the deviation between the two FIDE estimates. My analysis produces slightly higher estimates up to a Blitz rating of around 2,235; above this, the estimates are lower. Nonetheless, the deviations remain small, indicating a good overall fit.

5. Limitations & Outlook

This method of estimating the relationship between FIDE and Lichess ratings has several limitations. The most important is that the FIDE ratings are self-reported. They may be outdated, inaccurate, or fabricated.
The significant number of outliers supports this concern. In fact, 12.7% of profiles were excluded due to extreme differences between Lichess and FIDE ratings. Some users reported FIDE ratings of 3,000 while having Blitz ratings below 2,200. Others claimed FIDE ratings of 600 despite Blitz ratings above 2,300. The former are clearly fabricated (even Magnus Carlsen is below 2,900 FIDE), while the latter are likely incorrect entries.
Furthermore, profile information is voluntary. Some users may choose not to disclose their FIDE rating, possibly because they feel it is too low compared to their Lichess rating. This introduces a potential bias.
Nonetheless, the method presented here provides an additional way to explore the relationship between online and official rating systems. Further work is needed to expand the dataset and strengthen the reliability of the estimates.

References
[1] https://lichess.org/forum/general-chess-discussion/lichess-to-fide
[2] https://lichess.org/@/NoseKnowsAll/blog/introducing-a-universal-rating-converter-for-2024/X2QAH27t
[3] https://chessgoals.com/rating-comparison/