A plain-words guide
How to use EdgeSheet
This page explains every number on this product in everyday words. You do not need to know anything about betting or maths to read it. Start at the top and go down.
What this is, and what it is not
EdgeSheet is a study tool. It reads the odds from many betting sites, looks up what teams and players actually did, and does the maths for you. Then it shows you what it found.
It does
- Collect odds from many sites, every few minutes.
- Keep a record of real results — scores, stats, injuries, weather.
- Compare one site's price against all the others.
- Add up separate clues and say when they agree.
- Show its own past record, including when it was wrong.
It does not
- Take bets. There is no wallet and no stake anywhere here.
- Promise a win. Nobody can do that, and we never say it.
- Tell you how much to bet. We record no money at all.
- Know if betting is legal where you are. Check that yourself.
There are two very different kinds of edge
This is the most important idea on the page. Mixing these two up is the mistake that costs people money, so we keep them in separate boxes and never add them together.
1. A price that is cheaper
Imagine eight shops selling the same lamp. Seven say £20. One says £17. The cheap one is cheaper — you do not need to know anything about lamps to see it.
Betting sites work the same way. When one site offers a better price than all the others on exactly the same bet, that is a price opportunity. It is plain arithmetic. It needs no guessing about the game.
But it says nothing about who will win. A cheap price on a bad bet is still a bad bet.
2. Clues that point one way
Has the team been winning? Are the players healthy? Did they have a long rest? Is it windy? Each of these is a clue.
When several separate clues point the same way, we call that confluence and give it a score out of 100. Higher means the clues agree more.
Be careful here. We checked whether this score actually picks winners, and we could not find a link. See what we do not know.
The bar: the one picture to learn
Every bet has a score it must beat just to break even. Think of a high-jump bar. If the bar is at 52, you need to clear 52 to stay level. Clear more and you are ahead.
The price itself sets that bar. A price that pays a lot sets a low bar. A price that pays little sets a high one. That is how betting sites make their money.
- The thick line down the middle is the bar — the score the price needs.
- A bar stretching right means our measurement is above it. Good news.
- A bar stretching left means it is below. Not enough.
- A hatched stub in the middle means the difference is too small to call. We show no verdict, because the difference is smaller than our own measurements are precise.
The bar is drawn to fit the biggest difference on the page. So a full-width bar means “the biggest one here”, not “big”. The page says the scale under the findings.
Reading one finding, step by step
- What kind of finding it is. A cheap price, a market the clues point at, or one we looked at that did not pass. Always on the card, so you can never read one without knowing which it is.
- The bet itself, written the way a betting site writes it — Over 62.5 means “more than 62.5”.
- The bar from the section above. Our measurement against what the price needs.
- The clue score and how many separate clue groups had something to say. If we never scored it, this says not scored — never 0. Zero would be a score, and no score is not a score of zero.
- The price, and how many sites are offering this bet right now. We never call a price “the best” unless every site is on the same number — otherwise we would be comparing two different bets.
- What you can do. Open the market to see every site's price, how the number has moved, and what the bet would cost at a different number. Or keep it for later. Keeping a finding is just a bookmark — it does not place anything and records no money.
The six clue groups
A clue only counts if it is separate from the others. Two clues that are really the same thing twice would make us look more certain than we are.
We want six groups. Not every sport has all six — some need data we do not have yet. Here is the honest state of each:
How the team or player has been doing lately.
We have this for most sports.
How these two sides tend to go against each other.
We have this for most sports.
Injuries, and who is fit to play.
We have this for some sports only.
How many days off each side had.
We have this for most sports.
Which way the betting sites moved the number.
We have this for most sports.
Wind, cold and rain, for games played outside.
We have this for some sports only.
When a card says 3 evidence families, that is how many groups had real data. A market needs at least three before we will score it at all. If it has fewer, the card says so rather than showing a score built on one clue.
Putting the findings in the order you want
On Today's Picks you can change three things. None of them change which findings are there — only the order and how many you see.
Sort by
- Stats edge — how far our measurement from real data sits above what the price needs. This is the default, because it is what this product is for.
- Confluence score — how much the clues agree, for markets we scored.
- Most evidence — most separate clue groups first.
- Price edge — one site against the others. Plain arithmetic, and it says nothing about who wins.
- Starts soonest — by kick-off time.
Every one has a high to low and a low to highsetting. Pick whichever you need.
Day or week, and league
- This day shows one day, in your own time zone.
- This week shows the next seven days together.
- League keeps one competition only. It lists only leagues that actually have findings, so you can never pick an empty one.
One thing to know about sorting: some findings have no number for the thing you chose to sort by. Those go to the bottom, whichever direction you pick. We will not put “we could not measure this” at the top of a low-to-high list and let it look like the worst one. The page tells you how many those were.
Every market goes both ways
Most bets have two sides. “Over 62.5” and “Under 62.5” are the same market looked at from opposite ends. So are “home team wins” and “away team wins”. A football match has three: home, draw, away.
If our clues give Over a score of 58, then Under gets 42. They always add up to 100, because they are one measurement seen from two sides.
That is why every finding has The same market the other way underneath it. Open it and you see the other side with its own bar, its own score and its own price. You can click straight through to it.
Showing you only our side would be handing you an answer and hiding the comparison it came from. It also used to be the only thing stopping you from building the other half of a bet.
Words we use, in plain English
- Odds, or the price
- How much a bet pays. −110 means you risk 110 to win 100. +150 means you risk 100 to win 150.
- Break-even
- How often a bet must win just to stay level. A price of −110 needs to win about 52 times in 100.
- The line
- The number you are betting over or under. In "Over 62.5 yards", 62.5 is the line.
- Moneyline
- A straight bet on who wins. No line, no points — just the winner.
- Spread
- A head start, so the stronger side has to win by more than a set amount.
- Total
- A bet on the two sides’ scores added together, over or under a number.
- Prop
- A bet about one player, like how many yards they get, rather than about who wins.
- Confluence
- Our score, out of 100, for how much separate clues agree on one side.
- The vig, or the margin
- The betting site’s cut. It is built into the price, which is why break-even is above 50.
- De-vigged
- The price with that cut taken out, so two sites can be compared fairly.
- Stale
- A price we read a while ago that may have moved since. We show how old every price is.
- Closing price
- The last price before a game starts. Beating it is a sign of a good read, and we measure ourselves against it.
- An exchange
- A site where people bet against each other, not against the house. A price there might only be good for a tiny amount, and we cannot see how much.
The plain-English summary on each card
Every finding has a short summary under it, headed In plain English. It says the same thing as the numbers above it, in ordinary words.
What it is
- The card's own numbers, turned into sentences by a fixed rule.
- The same every time. The same card always gives the same words.
- Written to say less than the numbers, never more.
What it is not
- Not written by an AI chatbot. No language model reads these games or writes these words.
- Not extra information. It knows nothing the card does not show you.
- Not an opinion, a tip, or a reason to bet.
We built it this way on purpose. A computer that writes fluent sentences can be fluently wrong, and you would have no way to check it. These sentences come from a fixed rule instead. Every one of them traces back to a number on the card. So we can test that it never says something those numbers do not support.
It will never tell you a bet is worth taking when we cannot confirm the site's fees. Where that is true it says the figure is before costs, and stops there.
What we do not know — please read this
This is the section most guides leave out. It matters more than the rest of the page.
Our clue score has not been shown to pick winners
We kept 198 of our own past findings and checked them against the last price before each game. If our score were any good, higher scores would have done better. We looked for that link and could not find it.
Worse: the findings we rejected did slightly better than the ones we kept. That is not what a working score looks like.
So read a confluence score as “here is what our clues say, and here is our working”. Not as a tip.
Our win rate is not evidence of skill
Our past findings won a bit more than half the time. That sounds good and is not. Most of them were bets on favourites, which win often and pay little. Winning often at short prices is what backing favourites looks like — not skill.
Not enough results yet to judge anything
To say “this method works” you need many results, and they must be separate games. Thirty bets on seven games is not thirty pieces of evidence — the same quarterback having a bad day moves all of them at once. By that honest count, no method here has enough results yet.
Things we simply do not have
- News. We do not read news stories. A trade or a suspension reported this morning is not in any number here.
- Some sports have no player data. Baseball, hockey, football and tennis have no per-player numbers in this product.
- How much you could actually bet. We record no sizes. A price might be good for very little.
- Fees. For some sites we cannot confirm the charges. Where that is true, every figure we show is before costs, and we say so on the card.
- Whether you may legally bet. We list sites we read prices from. That is not a statement that any of them is available to you.
You can check all of this yourself. The method record shows our findings against the prices they were made at, and backtests shows how the engine held up on past games.
Rules this product keeps
- Never zero for unknown. If we do not know something, we say we do not know. A missing number is never shown as 0.
- Nothing is promised. You will not find a win described as certain here, and no finding is ever called a sure one.
- The two kinds of edge are never added up. A cheap price and strong clues are different things, in different boxes, in different words.
- Test data is labelled. If you are looking at practice numbers rather than real ones, a banner says so on every screen.
- Keeping a finding is a bookmark. It records what we said and when. It is not a bet and no amount of money is stored.
- We show what did not qualify. “We looked and it did not pass” and “we never looked” are different, so we count both.
Betting costs money and most people lose. Nothing on this product is advice to place a wager. If betting stops being fun, stop.