DOCUMENT MS-HOW-01

MINTSLIPS

EXHIBITS: HOW MINTSLIPS WORKS

SUBJECT: A MODEL, A HARNESS, AND A MARKET5 EXHIBITS

MintSlips publishes a chance for each side of a market, and our price: that same chance written in cents. That is the whole product. It does not trade.

The trading happens in software you build and run, which we call your harness. It can be an agent, a bot or a short script. It reads our chance, weighs it against your own inputs, and places orders on the prediction market from your account. Our chance is one input among several. The exhibits below show each piece, and where MintSlips stops.

  1. EXHIBIT A · THE PIPELINEDIAGRAM
  2. EXHIBIT B · ONE MARKET, ONE GAMEREPLAY
  3. EXHIBIT C · MESSAGE TRACELOG
  4. EXHIBIT D · THE HARNESS IS YOURSINTERACTIVE
  5. EXHIBIT E · CALIBRATION RECORDTABLE

Teams are fictional. Venue prices and fees shown in these exhibits are illustrative and marked with *.

EXHIBIT A

THE PIPELINE

Data comes in, our models turn it into a chance, and the chance goes out over the site, the REST API and the MCP server. From there it is your harness that does the deciding. It combines our chance with your own model, your bankroll rules and the live venue price it fetches itself, then trades from your account. Green marks MintSlips, blue marks your harness. Everything else is yours.

EXHIBIT B

ANALYSIS OF ONE MARKET

A fictional live game. Our live models re-price about every 30 seconds. Each new chance goes to your harness, which checks its own rule against the venue price. Most updates end in SKIP.

HARBOR CITY @ NORTH RIDGE · FOOTBALLLIVE
MARKET: HARBOR CITY DEFENSE 3+ SACKSSIDE: NO
HARNESS RULE: BUY 5 IF PRICE + FEE ≤ OUR CHANCE, MAX 2 ENTRIES
PREGAMESACKS 0 / 3
MINTSLIPS CHANCE (NO)VENUE PRICE* (ILLUSTRATIVE)YOUR HARNESS BOUGHTSACK
DECISIONS, ONE PER UPDATEB = BOUGHT · H = RULE SAYS YES, CAP REACHED · · = SKIP
MEASURED LATENCY, PLAY TO YOUR HARNESS:
PLAY→ DATA FEED ~30-45 S→ OUR CARD ~45 S→ API ~1-2 MIN

Our chance arrives after the play. It does not get there before the market. Your harness should treat it as a considered second opinion, not as a race.

EXHIBIT C

MESSAGE TRACE

The same moment from Exhibit B, as your harness logs it. It talks to two parties, MintSlips (over MCP or REST) and the venue (over your own account), and both conversations go through your harness. The venue never talks to MintSlips. Illustrative trace; fields trimmed.

MINTSLIPSYOUR HARNESS (LOCAL)THE VENUE, YOUR ACCOUNT
EXHIBIT D

THE HARNESS IS YOURS

Here is a sample harness written as a decision tree. MintSlips supplies one number per market and you write every check. Move the controls and the tree re-runs on six sample markets. Your model's numbers, venue prices* and the $1,000 bankroll are illustrative. Fee is a flat 2¢*.

SAMPLE HARNESS, MARKET 1 OF 6
SAMPLE MARKETS (CLICK ONE)
MINTSLIPS ANSWERS ONE QUESTION. check_price tells you whether a price plus the venue's fee is at or under our chance. Which threshold to use, how much to stake, whether to trust your own model over ours, and whether to trade at all are all decided in the code you write.
EXHIBIT E

CALIBRATION RECORD

A chance is useful only if it means what it says. For the sides we show, we compare how often they hit with how often we said they would. A dot right of the line means the sides hit more often than our chance, so the chance ran conservative. A dot left of it would mean we were overconfident. Results come from official game results. This record is public on the site.

MODELS ARE IMPROVED CONTINUOUSLY. Before any new model goes live, it is tested on games it has not seen. The backtest rows above come from that kind of test. The live sample is still early, so read its gap as provisional.