Leadoff Roadrunner

About

Leadoff Roadrunner is a program, not a person. It watches a visiting team losing early and asks one narrow question: is the live price lower than the share of real games that reached the same state, with the visiting team batting, and ended with that visiting team winning?

It exists because the roster had a hole in it. Comeback Cicada reads the same kind of state but buys only the home team, so on 22 September 2026 a live road position had no rule that could carry it. Cicada was right to refuse. This is the rule that reads the other side.

A road team trailing is not a mirror of a home team trailing, and the rule is not a copy of Cicada's with one word swapped. The visiting team bats first, so it cannot be behind in the top of the first at all, and the window opens an inning later than Cicada's. Its base rates are lower at every matched state: down two in the top of the third is 23.70 percent across 17,685 games, against 33.92 percent for a home team at the same node. That spread is close to the plain home-field spread of 8.09 points, so most of it is playing at home rather than batting last. The published research on batting last finds little or no advantage in it.

Published 23 September 2026. Its record starts empty, and it stays empty until a live board and a live scoreboard both satisfy the rule above. It did not trade the road position that exposed the gap. That one belongs to a desk writer who took it on judgement before this rule existed, and writing it back into an agent's record would make every number on this page worth nothing.

It is worked by hand. The exchange's market feed carries prices and nothing about the game itself, so a person reads the inning, the outs, the runners and the score, checks the table, and enters the pick under the same published rule.

The rule

Watch a live game where the away team trails. In the top of the second, third or fourth inning, read the away team's Kalshi ask and compare it with the Retrosheet base rate for that exact state, visiting team batting: deficit, inning, outs and runners. If the ask sits at least 9 points below that base rate and is 20 cents or less, buy the away team at the ask. Sell at the entry plus 5 cents. If the sell does not fill, hold to settlement. If the gap is under 9 points, no pick.

Worked by hand. This rule reads the inning and the score, so a person reads them from the live game and enters the pick exactly as the rule says. Every pick is graded the same way as the rest of the Lab.

How it works

• Operated by hand. The registry marks it manual, so the scheduled runner skips it before evaluating anything: the markets API carries quotes, not inning, outs, runners or score.
• The base rates are counts, not a model. The table is a verbatim subset of the “V” rows of Greg Stoll's probs.txt, the data file behind his Win Expectancy Finder, generated from Retrosheet play-by-play. A “V” row is the visiting team batting and its win column counts the games that visiting team went on to win. The file's sha256, its read date and its span of 1903 to 2025 across 195,573 games are recorded in tools/agents/baserates.py. The information was obtained free of charge from and is copyrighted by Retrosheet.
• Only states the rule can reach are embedded: visiting team batting, innings two through four, deficits of one to nine runs. A rate is returned only at 300 games or more. Nothing is interpolated, smoothed or filled in from a neighbouring cell, and nothing is borrowed from the home table.
• The window opens in the top of the second because it cannot open earlier. An away team has never trailed in the top of the first, since nobody has batted yet, and the Retrosheet file holds no such row. Cicada's best-known state, a home team down three in the bottom of the first, has no away counterpart at all.
• The state has to be complete and exact: top of the inning, inning 2, 3 or 4, zero to two outs, a readable runner mask, integer scores on both sides, and an away deficit above zero. A partial feed is a refusal, because a missing outs count read as zero would be an invented game.
• It finds the away market by name. Kalshi's rules text puts the visiting team immediately before the word “vs”, so that name parses exactly; it is the multi-word home name that gets cut short, which is why Cicada has to work by elimination and this one does not.
• Then the arithmetic: the ask must be 20 cents or less, clear the same 100 resting and 3 cent spread gate, have traded at least once, and sit at least 9 points under the base rate. The row carries the state read, the base rate compared against, and a sell target of entry plus 5 cents.
• The 9 point gap is 1 point wider than Cicada's 8, and the extra point is the fee. Kalshi's taker fee on 100 contracts is 0.52 cents a contract at 8 cents and 1.12 cents at the 20 cent ceiling, so a 9 point gap on the screen is 7.9 to 8.5 points once the entry fee is paid.

It will not estimate a base rate it does not hold: a state outside the table, a cell counted off fewer than 300 games, or a feed that answers only partly is a refusal rather than a number. It also refuses every bottom of an inning, every inning outside the second through the fourth, every away team that is level or ahead, and every ask above 20 cents or inside 9 points of the count. Declining is the ordinary outcome and firing is the rare one: on most nights it reads the whole slate and takes nothing.

What would prove it wrong

If the trailing away teams it buys on a 9 point gap go on to win at about the share the market priced rather than the share the Retrosheet count implies, the early price was not overreacting and the premise is wrong. The other way it fails is quieter: the count pools 1903 to 2025, and if the modern game has moved the away comeback rate away from the pooled figure then the gap is an artefact of era rather than a mispricing.

Why this rule

Tests whether a live in-game price overreacts to an early deficit on the road, which is the half of the question Comeback Cicada cannot reach: its rule buys only the home team, so an away team trailing has no rule reading it at all. The comparison has to be against the away count, never the home one. An away team down two in the top of the third has won 23.70 percent of the time across 17,685 Retrosheet games; a home team at the same node has won 33.92 percent across 20,362. This agent is the measurement. It is NOT a claim that road comebacks are underpriced, and the roughly 10 point spread between those two numbers is close to the plain home-field spread of 8.09 points rather than a penalty for batting first. Operated by hand for now, for the same reason Cicada is: the Kalshi markets API carries quotes and nothing about the game.