Strategy lab
ledger as of 2026-11-19 · 361 settled bets replayedReplays the paper ledger under simple betting rules and shows what each would have made. Every rule sees the same 361 graded bets (wins, losses, pushes) in game-date order; voids and ungradeable bets risked nothing and are left out. Only Season 1 bets, from Nov 15 2026 on, are replayed; older bets are on the Ledger under before Season 1.
Only sources flagged hot: +1.71u
The rule that would have made the most, against -15.79u for betting everything (+5.9% ROI on 29 bets vs -4.5% on 361).
What each rule would have made
net units · ROI · bets takenSingles only
no flags neededSkip parlays and teasers; keep every straight bet.
Bets314
Units-7.05u
ROI-2.2%
vs baseline+8.74u
314.0u risked · luck range -11% to +7% · worst drawdown 25.3u
-20-10+0+10+20Nov 15-7.0u-15.8u
Singles onlyBet everything
Only sources flagged hot
flags as known that dayBet only while the bet's source was flagged hot that morning.
Bets29
Units+1.71u
ROI+5.9%
vs baseline+17.50u
29.0u risked · luck range -24% to +35% · worst drawdown 4.4u
-20-10+0+10+20Nov 15+1.7u-15.8u
Only sources flagged hotBet everything
Hot-source days: 2026-11-16, 2026-11-18.
Skip anything flagged cold
flags as known that dayBet everything except bets whose source, sport, bet type or tipster was flagged cold that morning.
Bets325
Units-13.24u
ROI-4.2%
vs baseline+2.55u
317.5u risked · luck range -13% to +5% · worst drawdown 31.6u
-20-10+0+10+20Nov 15-13.2u-15.8u
Skip anything flagged coldBet everything
Bet against anything cold
flags as known that dayTake the other side of straight bets whose source, sport, bet type or tipster was flagged cold that morning.
Bets20
Units-4.91u
ROI-24.5%
vs baseline+10.88u
20.0u risked · luck range -55% to +12% · worst drawdown 5.8u
-20-10+0+10+20Nov 15-4.9u-15.8u
Bet against anything coldBet everything
The ledger has no price for the other side, so a fade is priced at the opposite break-even plus a standard 4.8% margin (fading -110 is -110; fading +200 is -250).
Bet against cold tipsters
no data yetTake the other side of straight bets from tipsters flagged cold that morning.
No tipster was ever flagged cold at the time: a flag needs 10 graded bets in 14 days, and the busiest tipster has 33 settled bets over 3 game days. This rule needs a few weeks of Reddit history before it can be tested.
Flat stakes vs sizing by confidence
confidence stored at postingThe bot's own picks and parlays carry a confidence score when they're posted (bot picks 55-83; bot (own research) 55-76; bot parlays (own research) 10-60; bot parlays 14-76; bot picks not posted 60-61). Sizing bets stake × score ÷ that source's average score, so the total risked is the same as flat and only the mix changes. The average is taken over the whole history, a fixed scale per source: it sets how big a unit is, not which bets get more.
Flat
Bets156
Units-1.84u
ROI-1.3%
vs baseline+13.95u
144.5u risked · luck range -13% to +11% · worst drawdown 14.4u
Sized by confidence
Bets156
Units-0.52u
ROI-0.4%
vs baseline+15.27u
144.5u risked · luck range -13% to +12% · worst drawdown 13.7u
-20-10+0+10+20Nov 15-0.5u-1.8u-15.8u
Sized by confidenceFlat (bot bets)Bet everything
156 of 361 settled bets have a score (Bot picks, Bot (own research), Bot parlays (own research), Bot parlays, Bot picks not posted); the other 205 (Ladder, Pick of the Day, Reddit threads) have none, so they are left out of this comparison.
Bets that beat the closing line
hindsight: close known at game timeSettled bets split by CLV: whether their price was better than DraftKings' closing price (average over the bet's legs with a close). The close is only known at game time, so this isn't a rule you could have bet; it tests whether beating the close goes with winning.
Beat the close
Bets40
Units-4.12u
ROI-11.3%
vs baseline+11.67u
36.5u risked · luck range -37% to +15% · worst drawdown 11.0u
Didn't
Bets63
Units-9.88u
ROI-17.2%
vs baseline+5.91u
57.5u risked · luck range -36% to +2% · worst drawdown 12.4u
-20-10+0+10+20Nov 15-4.1u-9.9u-15.8u
Beat the closeDidn'tBet everything
103 of 361 settled bets have a closing line; the rest are left out of this split.
Bot picks: Reddit-led vs own research
source recorded at postingThe bot's settled picks and parlays split by where they came from. Most days research scores what Reddit picked; on a day Reddit has no consensus (too few usable comments) the bot picks on its own from the slate and DraftKings' lines, and the post says so. Not a rule you could bet: it tests whether the bot's own research holds up next to Reddit's.
Bot (Reddit-led)
Bets100
Units+1.65u
ROI+1.8%
vs baseline+17.44u
93.5u risked · luck range -13% to +16% · worst drawdown 9.7u
Bot (own research)
Bets56
Units-3.49u
ROI-6.8%
vs baseline+12.30u
51.0u risked · luck range -28% to +14% · worst drawdown 6.3u
-20-10+0+10+20Nov 15+1.6u-3.5u-15.8u
Bot (Reddit-led)Bot (own research)Bet everything
How the replay works
- Hot and cold flags are the ones the rule would have seen that morning. For each game day the flags are recomputed from results through the day before, as graded, with today's tracking rules (14-day window, at least 10 graded bets, ROI at or above +20% for hot or at or below -20% for cold, or a 5-bet streak). Nothing from the game day itself or later is used; the one assumption is that the previous day was graded before the bet was placed (the grader runs early morning ET). The live bot did not store daily flags back then, so these are reconstructed, not the alerts that were actually posted.
- Charts run along the 361 bets in game-date order, so every rule shares one x-axis; a flat stretch means the rule sat those bets out.
- Same stakes as the ledger unless the rule says otherwise: 1u per straight bet, 0.5u per bot parlay. Units are net; ROI = units ÷ units risked; the luck range is the same 90% bootstrap as the readiness page.
- Small samples. 2 game days had a hot source. Rules that take few bets swing on a handful of results; read their luck range before their units.
- Missing for later: a few weeks of tipster history (for the tipster rules).