Altora Analysis
CS2 · LoL · Dota 2 — calibrated model analysis, not tips. Tracked in the open.
Model validation · 13,586 out-of-sample matches · to 2026-07-30
These are matches the model never trained on. It picks the winner 62% of the time and its probabilities are calibrated. On esports the market is sharper than the model — it sees team news (rotations, injuries, rested players) that a results-based model can't — so we show that gap openly rather than hide it.
Reliability — model probability predicted vs what actually happened
The bars line up on every band — the model is honestly calibrated. It isn't as sharp as the market on esports, and we say so. Analysis in the open, not tips.
Since 21 July 2026, Dota 2 matches are priced by a dedicated Dota model (rating built from 114,000+ professional games with margin-of-victory weighting). CS2 and League of Legends continue on the shared model. The validation numbers above pre-date this change; the Dota model's own record accrues in the live table below and will be reported separately once the sample is meaningful.
The live record · every pick since going public
Esports runs year-round (CS2, LoL, Dota 2). Nothing deleted — wins and losses, every one. The validation above is the model's accuracy over thousands of past matches.
| Date | Model pick | Prob | Result |
|---|---|---|---|
| 5 voided match(es) shown greyed below — kept on the record, excluded from the accuracy figures (no result to score). | |||
| 2026-10-17 | team falcons v bb team | 74% | L |
| 2026-10-16 | 100 thieves v sangal | 65% | W |
| 2026-10-16 | legacy v bb team | 65% | L |
| 2026-10-16 | cloud9 v furia esports | 81% | L |
| 2026-10-15 | karmine corp v cloud9 | 66% | W |
| 2026-10-15 | aurora gaming v ninjas in pyjamas | 73% | W |
| 2026-10-15 | team falcons v 9z | 70% | W |
| 2026-10-15 | furia v legacy | 50% | W |
| 2026-10-14 | aurora gaming v 9z | 59% | L |
| 2026-10-14 | furia esports v bb team | 68% | W |
| 2026-10-14 | team falcons v ninjas in pyjamas | 77% | W |
| 2026-10-14 | legacy v parivision | 64% | W |
| 2026-10-11 | spirit v aurora gaming | 70% | L |
| 2026-10-11 | vitality v mouz | 65% | W |
| 2026-10-10 | vitality v aurora gaming | 72% | W |
| 2026-10-10 | spirit v mouz | 59% | L |
| 2026-10-09 | aurora gaming v 1w team | 70% | W |
| 2026-10-07 | furia v 9z | 66% | W |
| 2026-10-07 | spirit v m80 | 87% | W |
| 2026-10-06 | team spirit v 1w team | 89% | L |
| 2026-10-06 | furia v aurora gaming | 56% | L |
| 2026-10-06 | legacy v m80 | 75% | L |
| 2026-10-05 | g2 v shinden | 80% | W |
| 2026-10-05 | legacy v 1w team | 54% | L |
| 2026-10-05 | natus vincere v 9z | 61% | W |
| 2026-10-05 | aurora gaming v bb team | 64% | W |
| 2026-10-04 | mouz v furia | 52% | W |
| 2026-10-04 | vitality v natus vincere | 73% | W |
| 2026-10-04 | bb team v m80 | 70% | W |
| 2026-10-04 | lyon v team liquid | 53% | L |
⬇ Download the full record (CSV — opens in Excel / Sheets) · showing the most recent 30 of 322
Model updated 22 July 2026: published probabilities now blend the Elo rating with the market opening price through a gradient-boosted calibration layer, validated out-of-sample before switching. The record below spans both model versions.