Removing them raised the record by 0.4 units.
Kick-off 21 Oct 2026
Kick-off 17 Oct 2026
Kick-off 16 Oct 2026
Kick-off 11 Oct 2026
Kick-off 11 Sept 2026
Kick-off 6 Sept 2026
Kick-off 19 Jul 2026
Kick-off 3 Jul 2026
Kick-off 7 Jun 2026
What each reason means
Logged after kick-off
Posted after kick-off
Removed on re-check
Voided after posting
Price not confirmed
Posted ↗
Changes to how we count
We no longer name bookmakers
From 13 October 2026 we no longer name bookmakers on the site, in the API or in new ledger files.
Read more
From 13 October 2026 we no longer name bookmakers on the site, in the API or in new ledger files. Earlier ledger files and seals keep the names they were published with: they are timestamped and we don't rewrite them.
Nothing in the record changes: every pick keeps its price and its result. Prices differ between bookmakers, so each pick shows the minimum price at which it still has value.
A match that restarts part-way is never picked as a new game
A stopped match restarts at the score it stopped at, so we no longer record new picks on it.
Read more
When a match is stopped part-way and moved to another day, it restarts at the score it stopped at, but our fixtures feed lists it as a new match that has not started. Our strategies could therefore treat it as a fresh 0-0 game. We now mark these matches, and our pick records refuse any new pick on them.
A pick that was voided because its match was postponed or stopped is now graded only if the match is played within 48 hours of its original kick-off; otherwise it stays void. This is the rule for picks made from 13 October 2026. Earlier picks keep the previous rule: graded if the match was played within 5 days of the pick.
Corrected 13 October 2026. Two picks from internal test strategies, made that morning on the restarting Sporting JAX v Birmingham Legion match, were voided before it resumed. Neither was published, and no published pick changed.
A pick is never graded before its match has been played
A postponed match was briefly reported as finished and one pick was graded lost three days early.
Read more
Sporting JAX v Birmingham Legion (USL Championship) was stopped at 0-1 on 10 October and moved to 14 October. We had correctly voided the pick on it. The next day our results feed briefly reported the match as finished, and the pick was graded as lost on the 0-1 score, three days before the game was due to resume. We now never mark a match finished, or grade a pick on it, while its kick-off is still ahead.
Corrected 12 October 2026. That one pick (home win at 3.06) is void again, so a €10 loss left the record. No other settled pick had been graded on a match that had not been played.
VIP picks are graded at the price we gave VIP members
Some VIP strategies record each pick at the second-best price; their records had been graded at the single best price.
Read more
Some VIP strategies choose and record each pick at the second-best bookmaker’s price, so that a member can take it at more than one bookmaker. Their records had been graded at the single best price instead, which was up to a few percent higher. We now grade them at the price we recorded and sent.
Restated 3 October 2026. VIP match-result record, as of that day (105 settled picks): ROI +13.4% → +12.0%, closing-line value +2.5% → +1.3%. 61 of those picks were re-graded, each one lower, never higher; 28 more (5 then unsettled, and 23 from a former VIP over/under strategy) were corrected the same way. Every pick’s seal still checks out: the sealed price is the higher one, and the correction is logged.
Prices no other bookmaker offered are re-graded, and the page leads with today’s strategies
A price more than 1.25× the sharp bookmaker’s with no other bookmaker within 5% is re-graded at the best price another bookmaker actually offered, and listed as “Price not confirmed”.
Read more
Every pick is graded at the best price available when it was made. We now check that price: if it was more than 1.25× the sharp benchmark’s price and no other bookmaker came within 5% of it, a reader could probably not have taken it. Such a pick is re-graded at the best price another bookmaker actually offered at that moment, and listed above as “Price not confirmed”. Picks for which too few prices from that moment are still stored (the sharp benchmark’s, or at least three other bookmakers’) are counted and shown separately, with a worst case.
Restated 3 October 2026. On that day the all-time record was +5.24% on 1,042 picks; with 5 picks re-graded at another bookmaker’s price it became +5.20% on the same 1,042 picks. 564 picks could no longer be price-checked, because too few prices from the moment of the pick are still stored; they are counted in the figure above. A deliberately pessimistic re-check of those — re-grading every one whose price was more than 1.25× the sharp bookmaker’s lowest stored price at that sharp price — gives a worst case of +3.9%. The current line-up’s figure did not change.
From that day the headline number was the current line-up — the strategies running then — because that is what you would follow; on 4 October 2026 it went back to the all-time record. At the time, more than all of the all-time profit came from one strategy we stopped on 2 October 2026 after finding that its edges came from single bookmakers pricing far above the sharp benchmark, some of them broken prices.
A closing price that stopped moving no longer counts as the close
If the sharp bookmaker’s last price was more than two hours old while the market moved, we use the exchange’s close or an adjusted price instead.
Read more
We grade every pick against the sharp benchmark’s last price before kick-off. Sometimes that price had not changed for hours while the rest of the market moved — so it was not really a closing price. Since 30 September 2026 we check this at kick-off: if the sharp benchmark’s price was more than two hours old and other bookmakers or a betting exchange had clearly moved away from it, we use the exchange’s closing price instead, or the sharp benchmark’s price adjusted by how far the market moved. If neither is available, the pick gets no sharp closing value. About 8% of picks are affected.
Restated 30 September 2026. We re-checked the last seven days of picks (older prices are no longer stored in enough detail to re-check). Three public strategies moved by more than half a point: goals over/under — two sources (+1.9% → +1.1%), bookmaker agreement, strongest tier (−0.1% → −0.7%) and the VIP late 1X2 line (+2.3% → +0.5%). Every change was downward for these three; results (ROI, wins, losses) do not change — only the closing-price comparison does.
Which closing-price comparison is shown first
/picks now compares each pick with the closing price at the sharpest bookmaker first, the same figure /performance shows.
Read more
/picks now leads each pick’s closing-price comparison (CLV) with the sharp closing line (the sharpest bookmaker’s closing price with its margin removed, or a 5+-bookmaker consensus), the same figure /performance shows. The older figure against the bookmaker’s own close stays second, labelled “vs book close”. The picks, the rules and the record are unchanged — only which figure leads.
Records that span rule changes are labelled
When we change a strategy’s rule its record is not restarted.
Read more
When we change a strategy’s rule, its record is not restarted: picks made under the earlier rule stay in its figures. Since 30 September 2026 such a strategy carries a small label on /performance — “Includes picks under earlier rules” (until 13 October 2026 with our internal version codes) — so you can tell a record of one rule from one that spans several.
The record counts every pick made while a strategy was Active
Until then retiring a strategy removed its whole history, which made the number look better.
Read more
Changed 29 September 2026. Until then the record counted only the strategies Active at the moment you looked: promoting one pulled in its Testing picks, and retiring one removed its whole history — our over/under strategy left the record with 252 picks at +4.2% when we retired it on 24 September, which made the number look better than it was. The all-time record now includes it and the other strategies that counted before they were retired (for the period before we labelled strategies Active, those that were in the published record then). The restated all-time figure is lower than the old headline.
Also from 29 September 2026: some strategies have a ceiling above which we never show or send their picks, because their high-edge picks tested worse. Picks above that ceiling are left out of the headline, because no reader could have followed them. They still count on the strategy’s own row.
The headline counts every pick logged before kick-off, many also posted to our Telegram channel (opens in new tab). The picks above are not in it, except the “price not confirmed” ones, which stay in, re-graded at the lower price. How the record is kept: methodology · full record