21 Blitz: How This Guide Was Derived

Why this page exists

This page documents where the guide’s rules came from, how they were tested, and which details remain uncertain. You can use it to check the research before or after buying the guide.

---

The simulator

I built a complete 21 Blitz engine covering all game mechanics:

• Four columns, three-strikes ending
• Wild Black Jacks (J♣ and J♠ only) that count as ten-value cards AND clear any column
• The stacking award logic: a five-card 21 pays 1,000 points (400 for the 21 plus 600 for the five-card bonus); a wild-assisted five-card 21 pays up to 1,200
• The streak schedule: 0 / 250 / 500 / 750 / 1,000 / 1,250 on consecutive clears, reset by any non-clearing placement or bust
• The 180-second clock at exactly 10 points per second remaining, paid even on a three-strike bust-out ending
• Bust behavior: the busting card is lost, the column clears, a strike is charged, and streak progress resets

Scoring validation: All per-clear award values and the bust model were validated against 28 real game-over panels read from recorded matches. Every panel reconciles exactly as:

Final = Base + Streaks + Combo + Time

The Base/Combo split on the game-over screen is a display convention. Five-card 21 clears land in the “Combo” line and ordinary clears land in “Base,” but the underlying per-clear values are the ones listed above. The 28 panels establish that result directly.

Placement timing: Each placement was modeled at 1.0 seconds; each undo cycle at 1.5 seconds (costing roughly 15 displayed score points). These are derived from observed play pace, not per-player measurement.

---

Key benchmark results

All strategy comparisons used paired identical deck orders. Every strategy faced the same card sequence, which removes deck luck from the comparison.

Recommended strategy performance

The strongest strategy tested was an expected-value search that uses the guide’s board-spread and escape-column principles. It was evaluated on 750 fresh decks that were not used for tuning:

Metric	Result
Average displayed score	~8,730
Average clears per game	14.8
Average busts per game	1.2

These are out-of-sample results on decks the strategy was never tuned against.

The five-card trap (120 paired decks)

The expected-value search was run twice on the same 120 decks to test the five-card premium:

• Premium-weighted: the search values five-card 21 clears at their posted 1,000-point payout (roughly 1.5x a normal 21)
• Equal-value: the search prices every clear identically, ignoring the five-card premium

Results:

Metric	Premium-weighted	Equal-value
Average clears per game	lower	+1.0 more
Average busts per game	higher	−0.25 fewer
Three-bust-out rate	39%	31%

Sign test on clears: p < 0.0000001. Even the search player did worse when it chased the five-card premium. The extra busts and lost board flexibility cost more than the bonus paid. The paid guide turns this result into a practical rule.

Undo pricing (320 decks)

Against a selective-undo baseline (roughly 2–3 undos per game, at genuine decision points only):

Undo behavior	Score impact
3× as many undos	−233 displayed points (p = 0.003)
Undo on roughly half of placements	−502 displayed points (p < 0.000000001)

The extra information never recovered its clock cost at any tested frequency.

How close to optimal are the guide’s rules?

The recommended strategy was audited against a deep exhaustive search on sampled positions:

• Median decision cost: 0 points (the most common outcome is the same move)
• Mean decision cost: about 7 points
• Decisions meaningfully suboptimal: about 2%

The rules in the guide are a human compression of a policy already near the playable ceiling. The remaining gap is execution speed and accuracy.

---

Limitations

• Base/Combo split for wild-assisted five-card 21s is a display convention I inferred from the 28-panel dataset rather than exhaustively enumerating. The per-clear totals are unaffected.
• Placement timing (1.0s / undo 1.5s) is derived from observed play pace, not measured per-player. If you are materially faster or slower, the undo break-even point shifts proportionally.
• Benchmark numbers are simulator results. Live scores also depend on reading speed, mistakes under pressure, opponent pool, and card variance.

---

The paid guide contains the five-rule strategy, worked board examples, card-tracking method, and practice drills.