Development and surprise
When does a hidden option justify delaying a visible build?
01Buy an option
02Wait for the window
03Convert to points
The decision at the table
Development cards create possibilities whose value changes with the table: a robber move, a burst of resources, a road opportunity, an army race, or concealed points. Buying a card also postpones other uses of grain and ore.
The mechanism to test
Compare development purchase and play timing with visible construction. Track publicly known deck depletion without pretending to know other hands. Treat hidden victory potential as a distribution. Consider whether delaying visible points changes targeting, but verify that effect rather than assuming opponents are fooled.
Proposed experiment
Compare fixed development-card bonuses, deck-aware valuation, and timing-aware valuation with award-race beliefs. Hypothesis: conditional valuation improves wins over indiscriminate purchases while reducing cards that never affect the outcome.
Record buy-to-play time, award transfers, robber impact, and unused development counts at game end. Use public timelines for exhibits. Separate directly observed plays from inferred hidden points and preserve the server’s visibility rules.
First ablation
Switching the held-card and knight terms off changed the contrast by -0.035 wins per game (95% interval -0.144 to +0.074); raising them by half through the development slider changed it by +0.031 wins per game (95% interval -0.070 to +0.132). Neither is distinguishable from zero. Details and every run ID are in the ablation report; this is a development-tier result on one lineup.
Third round: the timing switches under the learned leaf
The three timing switches were re-run against the learned tables leaf, which buys about twice as many development cards as the hand-written leaf, so card timing could matter more there. Measured as swapped pairs across development, confirmation, and population stages (192 boards per arm), the purchase-tempo rule loses −0.043 wins per game (95% interval −0.076 to −0.009), holding knights for the before-the-roll unblock loses −0.035 (−0.062 to −0.009), and the shaped knight schedule on the quarter blend is worth +0.001 (−0.016 to +0.017). The tables already time purchases and knight plays well, so the hand-tuned rules fight the leaf instead of patching it; every switch stays off. The timing-tables report has every pair, the diagnostics, and the browser reading; this is development-tier engine-arena evidence.
What could disprove it
Largest Army and Longest Road are contests with opponents, not guaranteed point purchases. Spending heavily to take an award briefly may be worse than a permanent build. Evaluate the full continuation.
Agent notes
Use the shared experiment design to freeze candidate versions, full lineups, budgets, sample size, primary contrast, and stopping rules before collecting evidence. This is a draft study brief, not a preregistration. No run IDs exist for this proposal.
The deployable policy reads only its own observation and recipient-visible events. The diagram is a conceptual schematic. Build new measured exhibits from retained artifacts using the visual publishing guide.
Consult the official rules for card timing and award eligibility; the authoritative server determines this arena’s legal commands.
All approaches · Player’s guide · Experiment program
Tracked investigations
| Updated | Investigation | Status | Finding and next step |
|---|---|---|---|
| 2026-09-09 | A maximal development-card appetite | Active | Round one, 256 deterministic games: contrast +0.031 wins per game (95% interval -0.070 to +0.132). See the ablation report. Next: Read the confirmation cohort on fresh seeds where registered; otherwise retest under the confirmed depth-3 candidate before changing defaults. Log 2026-09-09 |
| 2026-09-09 | Development card and knight terms | Active | Round one, 256 deterministic games: contrast -0.035 wins per game (95% interval -0.144 to +0.074). See the ablation report. Next: Read the confirmation cohort on fresh seeds where registered; otherwise retest under the confirmed depth-3 candidate before changing defaults. Log 2026-09-09 |
| 2026-09-11 | When to buy and play development cards | Complete | No. Three one-switch arms against the unchanged depth-2 control, 256 deterministic games each: single-seating contrasts of +0.145, +0.043, and +0.098 on boards 0-63 and +0.074 and +0.066 on boards 64-127, all the size of the turn-order seating term between adjacent search slots. The knight_play_value arm re-measured as a swapped pair on boards 64-127 corrects to +0.004 (95% interval -0.061 to +0.069); the pair's seating term is +0.070 (+0.005 to +0.136). The diagnostics show the mechanisms work (the tempo rule halves purchases, holding knights forgoes army awards) without winning. All switches stay off by default. Next: Retest contextual card valuation under the ntuple-leaf baseline if a card question opens there; any future two-search contrast in this arena is registered as a swapped pair. Log 2026-09-11 |
| 2026-09-11 | Development-card timing under the learned leaf | Complete | The two candidate rules lose under the tables and the leaf term does nothing on a usable blend. Swapped pairs, 512 games each, seating-corrected effects: buy_tempo -0.061 (development, refuted), -0.006 (confirmation), -0.053 (population); hold_knight -0.029, -0.025, and -0.051 (population, refuted); knight_play_value on the quarter blend -0.002, +0.000, +0.004 (all inconclusive), and +0.109 (+0.055 to +0.164) on the full blend's development pair, a configuration itself 0.2 wins per game below the tables alone. After-the-fact pooling over the 192 stage seeds: buy_tempo -0.043 (-0.076 to -0.009), hold_knight -0.035 (-0.062 to -0.009), knight_play_value (quarter blend) +0.001 (-0.016 to +0.017). All twenty runs completed 256 of 256 games. The blended leaf was fixed to honor knight_play_value, with a test. No default changes; no switch is worth the browser. Next: Nothing for these switches. If a blend becomes the default leaf, re-read knight_play_value at that blend weight; the effect scales with the hand-written share (zero at 0.25, +0.109 at 1.0 on one stage). Log 2026-09-11 |