Quant rigor. Retail terms.
Alpha Geist prices Kalshi contracts the way a market-making desk would, and surfaces the spreads to retail traders on a subscription. Real models, real backtests, real losing trades on the track record.
Why this exists.
I started building Alpha Geist because I kept watching prediction-market prices drift several cents away from what the underlying data actually said, and nobody was calling it. I started pricing contracts in a notebook for my own trades. The notebook became a backtest. The backtest became a live model. The live model became this.
My day job is shipping production ML and data systems at scale — the kind of work where you don't get to ship on vibes, because evaluation, observability, and measurable rigor decide whether the system survives contact with real users. That's the discipline that goes into every model on this site. Not a gut call from someone on Twitter. Not a "trust me, this works." A backtest you can audit, a track record that includes the losing trades, and a fair-value number with confidence intervals attached.
I trade every signal with my own capital. If a model degrades, I'm losing on it the same week you are — which is exactly the alignment a paid signal service should have, and almost none of them do.
Edge over hype.
A few principles we're not willing to compromise on, even for growth.
- 01
Numbers, not vibes.
Every claim on this site has a number behind it. If we can't measure it, we don't sell it. The track-record page is the source of truth — losing trades included.
- 02
Retail-only, forever.
We're not building an enterprise product. The Quant tier is the ceiling. Funds and prop shops can build their own; retail traders mostly can't.
- 03
No execution. Ever.
We surface the call. You take the trade. Alpha Geist will never touch your bankroll, hold your funds, or auto-execute. That's a regulatory line we don't cross.
- 04
Show the losses.
Every resolved signal is in the track record. Survivorship-clean backtests are the entire point of having a backtest. When a model has a bad month, we publish the bad month.
- 05
Skin in the game.
The founder trades the signals with personal capital. When a model's edge degrades, we notice the same week subscribers do — because we're losing on it too.