Found while reviewing #30 sim output. Pre-existing — not introduced by hole 5 redesign.
Current bot sink rates (200 trials, MAX_STROKES=15, +-20° aim noise):
| Hole |
Name |
Par |
Sink Rate |
| 1 |
Slingshot |
2 |
0.5% |
| 3 |
Funnel |
3 |
12.5% |
| 4 |
Conveyor |
4 |
20.0% |
| 5 |
Teleport Trio (post-redesign) |
4 |
56.5% |
Hole 1 at 0.5% is the most concerning — the off-axis magnet may be too strong, or the bot's aim-at-hole strategy is fundamentally incompatible with the magnet pull.
The bot is dumb (always aims at hole, no path planning), so these aren't direct proxies for human play. But hole 1 being a near-zero suggests either:
- Magnet strength too high for a par-2 intro hole
- Bot strategy needs to be teleporter/feature-aware (separate issue)
Action: spot-play these holes and decide whether to tune the geometry or accept that the bot is just bad.
Found while reviewing #30 sim output. Pre-existing — not introduced by hole 5 redesign.
Current bot sink rates (200 trials, MAX_STROKES=15, +-20° aim noise):
Hole 1 at 0.5% is the most concerning — the off-axis magnet may be too strong, or the bot's aim-at-hole strategy is fundamentally incompatible with the magnet pull.
The bot is dumb (always aims at hole, no path planning), so these aren't direct proxies for human play. But hole 1 being a near-zero suggests either:
Action: spot-play these holes and decide whether to tune the geometry or accept that the bot is just bad.