Adjusted EV Calibration Evidence: High-Confidence Packs Show -16.87% Edge

🔍 Find Live Pokemon Card Price Gaps Automatically → GapSense.uk

The average Adjusted EV edge across 68 high-confidence packs is -16.87%. In other words, the Pokemon TCG conventional wisdom — "high Claimed EV = buy" — flips negative once you validate it with sufficient trade data.

This isn't a hypothesis. This is GapSense's own pack_ev_calibrations table — 158 packs, snapshot as of 2026-04-24 23:00 UTC. We're publishing these numbers in full and putting our own methodology on trial with our own data.

Live source: GapSense methodology · Adjusted EV § Live Evidence · machine-readable JSON: live-stats.json. The live table is refreshed every ~30 minutes by the production scheduler — current numbers may differ from the snapshot below by a few percentage points, but the directional thesis (high-confidence negative, low-confidence positive but unreliable, medium hollow) is stable across runs.

📊 Real Data Across 158 Packs: 3 Confidence Tiers

Confidence	Pack Count	Avg Adjusted EV Edge	Signal
high	68	-16.87%	Don't buy
medium	1	+4.65%	The only convergence zone
low	89	+28.04%	Headline only — not trustworthy

Total: 68 + 1 + 89 = 158 packs. The breakdown is the conclusion.

🔍 Why Do High-Confidence Packs Show a Negative Edge?

Claimed EV — the simple product of a card's public market price times its pull rate — inflates easily when trade data on rare cards is thin. If a SAR sells just once at $300, that single transaction enters the calculation as the representative value for the entire population. That's the thin tape problem.

Packs with sufficient trade history (high confidence, 68 packs) have already had their "apparent advantage" priced in by the market. That's why Adjusted EV, after accounting for pack costs, sinks to an average of -16.87%. 83.8% of those 68 high-confidence packs show a negative Adjusted EV edge — this isn't an outlier-driven mean, it's the modal outcome. The market is smarter than you think.

The Time We Got It Wrong

To be honest — we thought "high confidence = thick data = better signal." We were prioritizing high-confidence packs on our watchlist. Three months in, we noticed it: our win rate was dropping.

The cause was simple. Thick data for us means thick data for everyone else too. We were hunting arbitrage in a place where information asymmetry had already disappeared. When we recalculated Adjusted EV properly: -16.87%. We were losing for exactly the right reasons.

⚠️ The +28.04% Trap: Don't Buy Low-Confidence Packs

The 89 low-confidence packs average +28.04% edge. "Isn't that the buy signal?" Not so fast.

This number is a sampling artifact. Right after a new set drops, when only 2-3 SARs have traded, a single outlier can move the population mean by 10% or more. +28.04% doesn't mean "the market is overlooking this" — it means "the market hasn't finished price discovery yet." A flashy headline — but not an entry signal.

What you should do with low-confidence packs is wait, not buy. Keep them on your watchlist until the trade count crosses the threshold and they graduate to medium/high.

The Uncomfortable Truth: Only 1 Medium Pack

Out of 158 packs, only 1 is medium (+4.65%). That's an honest weakness in the methodology. The zone where Claimed EV and Adjusted EV converge at a meaningful level is essentially hollow.

A thin convergence zone means buyable moments are rare. It's an inconvenient truth for traders hunting weekly pack arbitrage — but that's what the data says.

Could This Methodology Be Wrong?

Objection 1: "68 packs is a small sample. Is -16.87% just noise?" That's a fair point. The exact decimal will move with each calibration cycle (live JSON has read between -16% and -21% over the past week), and treating any single snapshot as precisely -16.87% is overconfidence. But the direction is robust: 83.8% of high-confidence packs come in negative regardless of which day you sample. The thesis isn't that the magnitude is precisely -16.87%; it's that the sign is stably negative.

Objection 2: "Aren't high-confidence packs skewed toward older, more liquid sets?" Exactly right. And that's precisely what Adjusted EV is measuring. Adjusted EV isn't a "fair lottery device" — it's a selection filter. High confidence = liquid = edge has evaporated. That correlation isn't coincidence. It's structural.

FAQ

Q. What's the difference between Claimed EV and Adjusted EV?
Claimed EV is a simple calculation: card market price × pull rate. Adjusted EV corrects for the outlier bias caused by thin tape (sparse trade data). For packs with sufficient trade history, Adjusted EV more accurately reflects market reality.

Q. Should I really ignore the +28.04% on low-confidence packs?
As an entry signal, yes — ignore it. But adding them to your watchlist is valid. When the trade count graduates to medium/high, re-evaluate at that point using the Adjusted EV as your benchmark.

Next Steps

This calibration table is refreshed every ~30 minutes by the GapSense production scheduler — the live values are at /methodology/adjusted-ev/live-stats.json. Check your portfolio right now — are you opening or buying packs that sit in the high-confidence, negative-edge bucket?

Cite as: GapSense, "Adjusted EV — Live Evidence," https://gapsense.uk/methodology/adjusted-ev#live-evidence (snapshot 2026-04-24 23:00 UTC).

🔍 Find Live Pokemon Card Price Gaps Automatically → GapSense.uk

🔍 Find Live Pokemon Card Price Gaps Automatically → GapSense.uk