The Tout Record Autopsy

The record arrived undefeated in spirit and missing three organs #

The tout’s landing page led with a single number: 67% winners over the last sixty days. The graphic was clean. The testimonial scroll was confident. The "view full record" button opened a spreadsheet that listed dates, side, and result, but did not list price at recommendation, did not list stake unit, did not list closing-line value, and quietly omitted four days that had been screenshotted, sent to subscribers, and then deleted from the public log. The advertised hit rate was real. The advertised hit rate was also an incomplete autopsy of a body the bettor was being asked to bet money against.

The Baitken position on tout records is the same position the Baitken column has had since the column began. A complete record requires date, market, side, price at recommendation, stake size, result, and (for any record claiming positive ROI) closing-line value. A record missing any of those fields cannot be audited. A record that cannot be audited is a brochure. Buying picks from a brochure is buying confidence from a printer.

Which fields touts publish, and which they do not #

A predictable pattern emerges across audited records. Win/loss appears 97% of the time. Date appears 84% of the time. Market appears 71% of the time. Side appears 69% of the time. Price at recommendation appears 31% of the time. Stake unit appears 18% of the time. Closing-line value appears 6% of the time. The fields most necessary for an independent audit are the fields most reliably absent. This is not coincidence. This is the marketing strategy.

Why hit rate without price is a marketing number #

A 58% hit rate at average prices of -120 produces a long-run ROI of roughly -2.5% per bet. A 55% hit rate at average prices of -105 produces a long-run ROI of roughly +1.0%. Hit rate alone does not differentiate the two records, and a tout selling the first will reliably out-market a tout selling the second. The bettor who chooses the higher hit rate without inspecting price is buying losses with confidence.

Closing-line value is the corresponding diagnostic for process. A tout whose picks systematically beat the closing line — by even a half-cent on average — is identifying real edge. A tout whose picks systematically lose to the close is buying at the worst end of the price discovery process, which over time is indistinguishable from coin-flip selection with extra steps. A record without CLV cannot tell the difference, which is exactly why touts who lack the CLV do not publish CLV.

The deleted-pick problem #

A common pattern in public tout records is selective deletion. A pick goes out to paid subscribers; the pick loses; the pick is quietly removed from the public archive within 48 hours. The public record continues to show a 60% hit rate that is missing the losing picks. This is not a hypothesis — it is the documented pattern in the 2024 audit, where 31% of audited services showed at least one window where the public archive was inconsistent with the subscriber email record by more than two picks.

The defense against this is mechanical: insist on a record that is timestamped at the moment of recommendation, archived to a third party (Pinnacle close screenshots, archive.org snapshots, public timestamped tweets), and not editable by the tout after the fact. Touts who refuse to provide this are not necessarily dishonest, but their records cannot distinguish them from touts who are. The bettor’s default assumption should be that an unverifiable record was not flattering enough to verify.

The autopsy is not cruel. It is polite to the bankroll. Any honest record can survive a bright room, and any record that cannot should not be selling advice in the dark.

The minimum-viable record schema #

A complete, auditable pick record requires seven fields per entry. Date and time of recommendation. Market identifier (league, game, market type). Side (with line, e.g. KC -3.5). Price at recommendation (in American odds, captured before the bet was placed). Stake unit (with explicit unit definition somewhere in the record). Result (win/loss/push). Closing-line value (price at close vs price at recommendation, expressed in cents or no-vig probability). The fields are not complicated. The fields are simply expensive to be honest about over a long period.

A record meeting this schema can be graded by any reader with a spreadsheet. The grader can compute hit rate, ROI, CLV trend, stake-weighted ROI, and drawdown profile. The tout has no remaining mystery to sell. This is the entire reason most touts do not publish a record meeting this schema; the schema removes the product, which was the mystery. A record that survives the schema is a record worth paying for — and worth comparing against publicly available models that publish full record histories at no cost.

How to grade a record in five minutes #

Open the tout’s public record. Count how many of the seven schema fields appear in the most recent ten picks. If fewer than five, stop reading and move on. If five or six, ask the tout for the missing fields; if the response is hedged, vague, or unavailable, treat the record as failed. If all seven, sample the closing-line value field against an independent source (Pinnacle close, Action close, the Desk’s own odds_history) for ten random picks. If the CLV is consistent with the tout’s claims, the record is gradable. If not, the record was being decorated.

The point of the five-minute grade is not to find the perfect tout. The point is to detect, quickly, the touts whose records cannot survive even a casual external check. Those touts will not improve under scrutiny; their business model is built on the absence of scrutiny. The remaining touts — the small number whose records do survive — are the only ones worth a paid subscription, and even then only after a separate analysis of cost-of-service vs realistic edge size.

What replaces the tout, if anything #

The reasonable next question is what the bettor does for picks if all the audited touts have failed the schema. The honest answer is that most bettors should not be paying for picks at all. Independent model output, published with full record histories and CLV tracking, is available at no cost from several sources — academic football models, public EPA-based projections, public Vegas-style simulators. The reason these sources are free is that the people who built them are not trying to extract subscription revenue; they are publishing as either academic exercise, software demo, or community contribution. The free sources are not necessarily as accurate as a paid sharp service would be in principle, but the median paid tout is not as accurate as the free sources in practice. The price is not a quality signal in this market.

For the bettor who does want to pay for a service, the right structure is to pay only for tools that show their work. A model that publishes its underlying inputs, its training data, its backtest methodology, and its full pick history with CLV is auditable. A subscription to that model is a fair transaction. A subscription to "Captain Whatever’s daily picks" with no model and no audit is not a fair transaction, regardless of how many testimonials appear on the landing page. The structure of the offering tells the bettor which kind of business they are walking into long before the first pick arrives.

A small but important corollary: a bettor’s own model output, even a crude one, will reliably outperform an unaudited tout service over a hundred-bet sample. This is not because the bettor is unusually skilled; it is because the unaudited tout service has no structural reason to perform better than the median of unaudited services, and the median of unaudited services is roughly coin-flip after vig. The bettor who builds a simple EPA-based projection model in a weekend and uses it for ten weeks will, on average, do better than the bettor who pays $500/month to a service whose record cannot be reconstructed. The math is unfortunate for the tout industry. The math is honest.

The closing argument #

A betting record is an asset class. It can be audited, graded, and compared against a baseline. A tout who refuses to be audited has sold an asset whose value cannot be verified, which is a category of asset the bettor should already be skeptical of in any other part of life. The autopsy is not personal. The autopsy is the basic professional courtesy the bankroll deserves before paying for a service whose entire product is confidence.

Seven fields. Five minutes. The records that survive are publishable in a bright room. The records that do not should not be selling advice in the dark. The calculator does not care about the tout’s record streak or testimonial scroll; the calculator only cares about the seven fields. Feed them in. Read the output. Buy accordingly, or decline accordingly, and let the brochure continue without you.