AI Sneaker Store

After 110 days, the AI’s store in Fontana, CA is up $680; the one-line rule’s store is up $330 and a store buying random releases is down $1,873.

Headline counts complete days only, to Sep 18. Counting the 31 provisional days to Oct 19, the three stand at up $1,340, up $878, down $2,307 (AI, rule, random).

A simulation on recorded GOAT and StockX sales with play money, not investment advice. The practice season (Jun 1 to Oct 18) was replayed from past sales; the live season (from Oct 19) is locked each day before that day’s sales are loaded. Days after the data’s complete date (Sep 18) are provisional and may change as the data fills in.

Why Fontana

Fontana, CA has the most GOAT resale sales per resident of the 291 US towns of 100,000 to 1.5 million people: 37.1 per 1,000 residents (36.3 to 37.9, 95% interval), from 8,664 sales shipped to a buyer there in the 365 days to Oct 11, 2026.

Method: only GOAT records buyer cities, and only on 256 of the 365 days, so the ranking rests on 2,516,660 located sales of 4,694,149. The rate divides by the larger of the Census place population and the residents of the ZIP codes carrying the city name. In this first version the town sets the sign and the demand picture above; it does not change which pairs the stores buy (they pick by release), and sizes are not tracked yet. The pick is reviewed every quarter.

What we learned before opening

The page exists because of two backtests on recorded sales, both run on October 19, 2026. They are why the stores buy the way they do. Neither is a forecast.

• Buying on the market lost money in every week. Buying pairs on StockX at the going price and reselling them lost about 20-25% per round trip whatever was picked, because fees and shipping are bigger than the price moves. Across 73 weekly groups at each horizon, no strategy had more than one profitable week. So the stores do not buy on the market at all.
• Buying every release at retail loses about 18-25%. Across 5,689 releases from February 2025 to June 2026, 3 pairs of each at retail plus 8% tax and $10 shipping lost about 18% per release on pairs that sold in 30 days, and about 25% when unsold pairs were sold later at the middle estimate. Only 6.5% of all releases made a profit.
• The price before release day predicts the result. Many shoes already trade before release. The pre-release median price divided by retail had a rank correlation of 0.70 with the 30-day result across 1,200 releases that had 5 or more sales to look at.
• The AI so far does no better than a one-line rule. “Buy the 2 releases a week with the highest pre-release price over retail” returned +48% per release at 30 days (interval +24% to +70%); the AI’s top 2 returned +39% (+22% to +57%). The AI was behind the rule at every horizon, not by a margin that proves anything. That is why the page runs both, and a random store, side by side.
• Getting the pair is the real constraint. The winners were the hyped drops, which are the hardest to buy at retail, and the backtest assumed it always got them. Here each retail buy goes through a simulated raffle with a stated win chance that falls as the pre-release premium rises (50% under 1.2 times retail, 20% from 1.2 to 2 times, 5% above 2 times). Those odds are an assumption, not a measurement, so the till shows profit with the odds and what it would be if every raffle was won.

How the strategy changes

• v1, from Jun 1, 2026. First version: AI vs one-line rule vs random, retail on release, simulated raffle. Why: the retail backtest: buying every release loses; pre-release median / retail predicts 30-day results (rho 0.70); the AI did not beat the one-line top-2 rule. Result so far, complete days to Sep 18: ai store up $680, rule store up $330, random store down $1,873.

Data: state, ledger, daily, log, strategy versions. Fees are the ones on the fee calculator. Related: AI Lab, price checker.