Rating targets
A target you
can verify.
In AccelaStudy® AI Chess, a rating target always names its rating pool, and finishing one means a rating recorded by the provider, not our guess.
This is how that is planned to work.
In development. Nothing on this page is a measured result.
Which rating
Every target names its pool.
You will see “1,600 Lichess rapid,” never a bare “1,600” or an unlabeled “Elo.”
A pool is a site, a variant, and a time control
A rating means something only inside the system that produced it. Each target will carry the rating provider, the chess variant, the time-control category, the rating system, and the date it was observed.
Pools are not interchangeable
Lichess rates players with Glicko-2 and FIDE rates them with Elo, and the two pools hold different players. There is no fixed conversion between them, so we do not offer one.
We start with one pool
The first target pool is Lichess rapid. That is a starting default, not a statement that the curriculum has been calibrated to it.
Other pools need their own work
Other time controls and other providers would each need their own data connection, eligibility rules, and validation before they could be offered. None of that is promised here.
Lichess is an independent organization. AccelaStudy AI Chess is not affiliated with or endorsed by Lichess.
What the numbers mean
Three records.
Kept apart on purpose.
Doing well in a course, being predicted to reach a target, and actually holding a rating are three different facts.
Learning progress
What you have worked through and what still needs review, based on how you did, what help you used, how well it stuck, and whether you could transfer it to a new position.
AVIAN readiness estimate (planned)
A prediction of whether you can meet a particular target, shown with its uncertainty and the evidence behind it. It stays absent until the model has been calibrated against independently verified players. Today there is no estimate.
Verified playing rating
The rating actually recorded by the provider for an account we have confirmed is yours, with its pool and date. Only the provider assigns it.
Self-reported is not verified. A rating you tell us is labeled self-reported and can never satisfy a verified graduation. An old best rating keeps its date and is not shown as your current one after a long break or a drop.
The states
Where you are in a course.
A course moves through these, and it can step back.
Weak retention or what shows up in your games can send you back to learning.
Not started
You have picked a target in a named pool and have not begun.
Placement
A short, resumable check of what you already show.
Learning
Instruction, practice, review, and transfer work toward the target.
Ready for verification
The learning requirements are met, and a verified rating is the open question.
Target verified
The provider’s rating and your play meet the target, and the record says exactly what was checked and when.
When something is missing, it says so
- Not enough evidence yet.
- The rating source is disconnected.
- The last rating we saw is stale.
- An assessment is unavailable.
- If you change your target, everything you have already done is kept and a new target record begins.
Graduation
What graduating at a target takes.
A verified result at a named target means the outcome was demonstrated, not that a lesson list was finished.
- An authenticated, non-provisional rating at or above the target, in the same pool.
- A recent sample of completed rated games in that pool, including play after the course’s final transfer assessment. How many games, and over what window, is still being validated; the policy will be published before graduation opens, and it is ours, not a Lichess rule.
- The course’s own learning requirements as well. One rating does not prove you finished the intended curriculum, and if you are already above the target you can show those requirements through placement and unseen positions.
- Games that do not count: aborted, unrated, wrong-variant, wrong-pool, duplicated, and games played with disallowed assistance.
- A provisional rating cannot graduate you, even if its first value is above the target.
A graduation record states the target, the pool, the policy version, the evidence window, and the date it was verified. Graduation is planned and is not enabled until its policy has been validated.
Hard limits
What it will never do.
These are not features we are saving for later.
Assign you a rating
Only a rating provider does that. We can show what you have learned and what the provider has recorded, and we keep those two things apart.
Turn a puzzle rating into a playing rating
A puzzle’s difficulty keeps its source label. Solving hard puzzles is not evidence of how you would do against an opponent, so no conversion is applied.
Guarantee a gain or a schedule
No model can make every learner reach an outside rating by a fixed date. What we can promise is narrower: graduation at a named target requires that demonstrated outcome.
Show a readiness number it cannot support
Until the estimate is calibrated, it stays absent rather than showing a number that sounds sure.
Honest reporting
We will report what happens,
including when it does not help.
The mapping from what AVIAN measures to a rating in a pool has to be earned with evidence.
- Calibration is planned with a consented pilot of players whose ratings are independently verified, using unseen positions and delayed retests, with training and test groups kept separate.
- We plan to check whether the course actually improves outcomes against a suitable comparison group, with the analysis written down before the results are seen.
- Until then, targets and observed gains are described as exactly that. We do not claim the course caused them.
- If learners improve, we say so. If they do not, we say that too, and we change the course.
Before you ask
Questions about targets.
Is my puzzle rating my chess rating?
What if I am already above my target?
What if I do not reach my target?
Will it work with a FIDE rating?
We will tell you when there is something to try
AccelaStudy AI Chess is in development. Join the waitlist and we will let you know as it takes shape.
Join the waitlist