What is the play score?

Play score is a 0-100 composite that ranks grading-flip opportunities on /plays/. It combines five inputs: net profit (40% weight), ROI (25%), comp depth (15%), pop-velocity (variable), and liquidity (variable). Higher scores mean stronger conviction, not just bigger profit.

How is play score different from the GRADE/WATCH/SKIP verdict?

The verdict is binary-ish — does the math support grading or not. The score is a continuous rank — among all GRADE-verdict cards, which ones have the strongest combination of size + return + data quality + tradeability. A $50K-net flip on a 1-month-old chase card with deep sample size beats a $50K-net flip on a vintage card with thin comp depth, even though net is identical.

Why isn't net profit the only factor?

Pure-profit rank gives you a list of vintage Charizards every time — the cards with the biggest dollar swing. But many of them have thin sample sizes (3-5 sales/year) and illiquid markets. Splitting out ROI rewards small-cost flips with proportional upside; depth rewards data quality; liquidity rewards faster-selling cards. The composite ranks flips by execution-realism, not just headline math.

What is the eligibility filter?

A card appears on /plays/ only if: net profit ≥ $50, ROI ≥ 30%, sample size in last 90 days ≥ 5, and pop velocity ≤ 75% (or null). This filters out cards where the math is technically positive but the data is too thin or the population is growing too fast to trust the slab premium.

How the Play Score is Calculated

The /plays/ index ranks the platform's grading-flip opportunities by play score — a 0-100 composite that combines net profit, ROI, data depth, pop velocity, and liquidity. This page documents the formula end-to-end. No black-box weights, no proprietary magic — every component is explicit and verifiable.

The formula

The composite, defined in factory/scripts/compute_plays.py:

play_score = profit_score + roi_score + depth_score + pop_score + liquidity_score

Maximum 100. Component caps:

Component	Max	Formula
Profit score	40	`min(40, 40 × log10(max(1, net_profit)) / log10(1000))`
ROI score	25	`min(25, 25 × roi_pct)`
Depth score	15	`min(15, 15 × sample_size_90d / 30)`
Pop score	~10	Varies by pop velocity (see component breakdown)
Liquidity score	~10	Varies by raw price band (see component breakdown)

Component breakdown

Profit score (40 points)

Logarithmic — diminishing returns above $1,000 net profit. A $50 net gets 18 points; a $500 net gets 36 points; a $5,000 net gets the full 40. The log scaling prevents a single $100K Charizard flip from drowning out a basket of $200 flips.

ROI score (25 points)

Linear up to 100% ROI. A 25% ROI flip gets 6 points; 50% ROI gets 12.5; 100%+ caps at 25. Rewards small-cost-base flips with proportional upside.

Depth score (15 points)

Sample size in the last 90 days. A card with 30+ sales gets the full 15; 15 sales gets 7.5; 5 sales (minimum for eligibility) gets 2.5. Rewards cards where the comp prices are statistically meaningful, not derived from 3-data-point series.

Pop score

Pop velocity = how fast the PSA 10 population is growing. A card with falling pop-velocity (graders running out of clean copies) earns more score; a card with accelerating pop-velocity (everyone grading at once) earns less. Caps at ~10 points.

Liquidity score

Cards in the $50-$1,000 raw band trade fastest — buyers exist at every price point. Sub-$30 cards have higher friction (eBay fees eat margin); $5K+ cards have thin buyer pools. The score rewards the liquid middle.

Eligibility filter

A card appears on /plays/ only when:

net_profit ≥ $50
roi_pct ≥ 0.30 (30%)
sample_size_90d ≥ 5
pop_velocity_30d_pct ≤ 0.75 OR NULL

Cards that fail eligibility might still appear in the optimizer with a verdict but won't surface as a play. The threshold isn't "is this a flip" — it's "is this a flip I'd recommend you actually execute given current data quality."

Worked example

Sheoldred, the Apocalypse (Dominaria United #290 normal):

Net profit: $111 → profit score = 40 × log10(111) / log10(1000) = 40 × 0.682 = 27.3
ROI: 38% → roi score = 25 × 0.38 = 9.5
Sample 90d: 22 → depth score = 15 × 22/30 = 11.0
Pop velocity: 0.45 (stable) → pop score ≈ 6
Raw $290 → liquidity score ≈ 8

play_score = 27.3 + 9.5 + 11.0 + 6 + 8 = ~62

Score 62 places it in WATCH-territory — solid but not GRADE-conviction. A card with $1,500 net + 80% ROI + 30 sample + falling pop + liquid raw price hits 80+ which the composite designates GRADE-conviction.

Frequently asked questions

What is the play score?: Play score is a 0-100 composite that ranks grading-flip opportunities on /plays/. It combines five inputs: net profit (40% weight), ROI (25%), comp depth (15%), pop-velocity (variable), and liquidity (variable). Higher scores mean stronger conviction, not just bigger profit.
How is play score different from the GRADE/WATCH/SKIP verdict?: The verdict is binary-ish — does the math support grading or not. The score is a continuous rank — among all GRADE-verdict cards, which ones have the strongest combination of size + return + data quality + tradeability. A $50K-net flip on a 1-month-old chase card with deep sample size beats a $50K-net flip on a vintage card with thin comp depth, even though net is identical.
Why isn't net profit the only factor?: Pure-profit rank gives you a list of vintage Charizards every time — the cards with the biggest dollar swing. But many of them have thin sample sizes (3-5 sales/year) and illiquid markets. Splitting out ROI rewards small-cost flips with proportional upside; depth rewards data quality; liquidity rewards faster-selling cards. The composite ranks flips by execution-realism, not just headline math.
What is the eligibility filter?: A card appears on /plays/ only if: net profit ≥ $50, ROI ≥ 30%, sample size in last 90 days ≥ 5, and pop velocity ≤ 75% (or null). This filters out cards where the math is technically positive but the data is too thin or the population is growing too fast to trust the slab premium.