Combat simulator
How often the odds are right
The simulator behind the overlay's odds, checked on real combats: for each combat it gives the chance of a win, a tie and a loss, and the game's log says what really happened.
Engine 0.8.0 · measured 12 October 2026 · 10,000 trials per combat
Solo
19 combats with the game's result
Most likely result was right
95%
18 of 19 combats
Average chance given to the real result
94%
over 19 scored combats
No odds: a card is not modelled
0%
Odds in every Solo combat with a result (19): no card was missing.
Brier score
0.078
0 is best, 2 is worst
Refused by the board check
0
Predicted and real
Every chance the simulator gave in Solo (a win, a tie and a loss per combat), grouped by what it predicted.
Read more: Predicted and real
Beside each group, how often that result really happened. When the two bars are close, the odds can be trusted.
Show as a table: Predicted and real, Solo
| Predicted | Predictions | Average predicted | Real |
|---|---|---|---|
| 0–20% | 36 | 0% | 0% |
| 20–40% | 2 | 33% | 50% |
| 40–60% | 1 | 54% | 100% |
| 60–80% | 1 | 74% | 0% |
| 80–100% | 17 | 100% | 100% |
Duos
232 combats with the game's result
Most likely result was right
82%
184 of 224 combats
Average chance given to the real result
81%
over 224 scored combats
No odds: a card is not modelled
3%
No odds in 3% of Duos combats (8 of 232): a card is not modelled.
Brier score
0.274
0 is best, 2 is worst
Refused by the board check
0
Predicted and real
Every chance the simulator gave in Duos (a win, a tie and a loss per combat), grouped by what it predicted.
Read more: Predicted and real
Beside each group, how often that result really happened. When the two bars are close, the odds can be trusted.
Show as a table: Predicted and real, Duos
| Predicted | Predictions | Average predicted | Real |
|---|---|---|---|
| 0–20% | 416 | 2% | 6% |
| 20–40% | 30 | 31% | 50% |
| 40–60% | 20 | 51% | 30% |
| 60–80% | 17 | 70% | 53% |
| 80–100% | 189 | 98% | 89% |
Why the odds missed
43 combats where the most likely result was not the real one, played on every board as it fought.
Read more: Why the odds missed
A card the simulator does not know counts as a plain minion. Largest cause first; Solo and Duos counted together.
| Cause | Combats | Share | Solo | Duos |
|---|---|---|---|---|
| Card played approximatelyNo replay was saved, and a card is played with a known gap. | 19 | 44% | 0 | 19 |
| Not knownNo replay was saved and every card is exact, so the cause cannot be told. | 11 | 26% | 1 | 10 |
| Not seenThe replay parts from the game, and a board brings what the log does not show (Blood Gem upgrades, a hand summon). | 7 | 16% | 0 | 7 |
| ChanceThe replay follows the game to the end: the odds were right and the less likely result happened. | 5 | 12% | 0 | 5 |
| Card not modelledNo replay was saved, and a card the simulator does not know was played as a plain minion. | 1 | 2% | 0 | 1 |
Cards that block the odds most
When one of these is on a board, the simulator gives no odds instead of a wrong number. Solo and Duos counted together.
| Card | Card id | Combats blocked |
|---|---|---|
| Stitched Salvager | BG31_999 | 7 |
| Ghastcoiler | BGS_008 | 1 |
Supported game versions
Newest build checked: 253216, on 12 October 2026.
The parser and the simulator were checked on games played on these builds.
| Build | Games |
|---|---|
| 253216 | 41 games |
What is measured, and its limits
Only the combats of the person who builds Tavern Ledger, from their own games, with the boards as they entered each combat (the first attack). Small numbers, one player: a guide, not a promise.
Read more: What is measured, and its limits
Start of combat effects are off, as in the overlay. A combat counts as scored when the simulator could play it; the share of combats with no odds is always shown beside it.
These numbers are refreshed after each round of new cards.