How to read a Ghost call
What a chance on SoccerGhost means, how it turns into odds, how it differs from a bookmaker's price, what makes a value pick, and how to check whether the Ghost has been right, with the arithmetic shown so you can check each step.
A chance, not a promise
Every number the Ghost gives is a chance. "Home win 62%" means that in matches like this one, priced like this, the home side should win about 62 times in 100. It also means the home side should fail to win about 38 times in 100.
So a single 62% call that loses does not show that the Ghost was wrong. A long run of 62% calls that came in only half the time would. That is why every call is kept in the Ghost Record, the losses included. The only fair test of a forecaster is a large number of its calls, set beside what it said.
From a chance to odds
To turn a chance into odds, divide one by it:
| The Ghost's chance | Fair odds |
|---|---|
| 80% | 1.25 |
| 62% | 1.61 |
| 50% | 2.00 |
| 25% | 4.00 |
These are fair odds. If the chance is right, backing the selection at that price would break even in the long run. A bookmaker builds a margin into its prices: the chances its prices for one market imply (one divided by each price) add up to more than 100%, so taken together they are shorter than fair.
The Ghost's number and the bookmakers'
The Ghost's chances come from football alone: results, the chances teams create and allow, recent form, the line-ups and where the match is played. Bookmakers' prices are never part of the calculation. They are shown beside the Ghost's numbers so that you see two separate opinions, not one repeating the other.
When SoccerGhost sets the two side by side, on the record and in its value picks, it first takes each bookmaker's margin out of its prices. What is left is the market's own fair view of each outcome, which can be compared with the Ghost's chance like for like.
What a value pick is
A value pick is not the Ghost's favourite result. It is a price that beats the bookmakers' own fair view of the same outcome. All of these must hold:
- The market is one that can carry a pick. Clean sheets, exact scores, corners, cards and shots never do.
- At least three bookmakers have priced the whole market, so their fair view is a middle value that one stale or misread price cannot set.
- The market gives the outcome at least a one-in-five chance. Long shots are where a stated edge is least reliable.
- Another bookmaker offers a price too, and the best price stands no more than 25% clear of the next bookmaker's. A price far beyond every other is what a bet the bookmaker later voids looks like.
- The best price on offer is at least 2% longer than the fair odds the market's view implies.
- The Ghost's own chance agrees that the price is worth taking: the chance times the price comes to at least one. For this test only, its chance can be lowered first, in the two cases below, and is never raised. The chance the site shows stays as it is.
- Every price the pick rests on was seen in the last two hours and passed the site's checks against misread prices.
Where any of these fails, there is no pick, however long the price.
Before the Ghost is asked to agree, its chance is lowered in two cases:
- On the low side of a goal line, such as under 2.5 goals or "no" to both teams scoring, two percentage points come off it: a 50% chance is tested as 48%. The club model expects slightly fewer goals than come in. The other models' chances are lowered the same way, as too few of their matches have been measured to show whether they lean.
- Where its record in that market has fallen short. Where the value picks the same model has made in the same market have come in less often than it said, by more than luck allows, its chance is lowered toward that record, the more so the longer the record.
How to check whether the Ghost has been right
The record gives two checks.
Does it come in as often as it says? Group the calls by what the Ghost said. Of the calls it gave about 60%, about 60% should have come in; of those at about 80%, about 80%. A forecaster whose 80% calls come in 65% of the time is overconfident, however many of them win.
Is it more accurate than the market? The Brier score measures this. For each call, take the chance given (0.62 for 62%) and what happened (1 if it came in, 0 if not), and square the difference: a 62% call that came in scores (1 − 0.62)² = 0.144, and one that lost scores 0.62² = 0.384. Averaged over many calls, lower is better. Always saying 50% on a two-way question scores 0.25. On the calls the bookmakers priced at the close, the record sets the Ghost's Brier score beside the market's and says which of the two, if either, has been the more accurate.
What no number does
Every chance on this site, the free picks and the hot bet included, is an estimate, never a guarantee. A 70% call loses about three times in ten, and a run of losses happens to every honest forecaster. SoccerGhost takes no bets and holds no money. If betting has stopped being fun, help is here.