This site counts draws — and everything that follows from not drawing
Roughly a quarter of all football matches end level. That single fact is duller than it sounds, because the draws do not arrive evenly: a team can play forty matches in a row without one. FDB4 holds more than half a million played matches, and most of the site is built to answer one question about them — how long do teams go without drawing, and does that mean anything?
The honest short answer, which the rest of the site then shows you in detail, is mostly no. A team on a long run without a draw is not owed one. That is not a disappointing result to be buried at the bottom of a page — it is the finding, and two of the tools below exist purely to measure how strongly it holds.
Start with the plain counting, because everything else on the site is measured against it. A no-draw run is the number of matches a team has played in a row without ending one level. A draw resets it to zero, and so does a hole in our coverage longer than 140 days — if we are missing matches, one of them could have been the draw, and a run we cannot vouch for is not a record.
Current streaks is the live version: who is on the longest run right now, filtered by competition or team type. It is the page to open if you want to know what is happening this week.
The records leaderboard is the all-time version, and the more interesting one: the longest runs ever recorded here, each with the team's own draw rate beside it so you can see how unlikely the run was for that team. Every row opens the run's own page — the actual list of matches that made it, with the league position the club held on the morning it started and on the morning it ended. That page is where a record stops being a number and becomes a season.
A run of thirty is only remarkable if thirty was unlikely, and you cannot say that without a number for how likely a draw was in each of those thirty matches. That is what the models are for. None of them predicts a result; each one turns a match into a probability so the counting above has something to be compared with.
Draw by goals is the explanation of how that is done at all: work out how many goals each side tends to score, turn that into the probability of every scoreline, and add up the level ones. The page sets the idea against this database's own record of every score from 0-0 upwards, and shows exactly where the textbook model gets the low scores wrong — which is precisely where draws are decided, and precisely what the Dixon-Coles correction of 1997 adjusts.
The draw predictor ranks teams by "draw debt" — the team's own draw rate multiplied by the length of its current run. Read the name carefully: it is a statistical expectation, not an obligation of reality. Nothing is owed. The ranking is useful for finding candidates to look at, and that is all it claims.
The Cross predictor does the same for a fixture rather than a team, by adding both sides' figures together — two teams each overdue on their own reading are a more interesting match than one. It needs a sign-in.
Most sites that count streaks stop before this part. These two pages exist to take the excitement back out of a record, and they are the ones worth reading if you only read one thing here.
Luck or skill asks whether a record is a feat or a coincidence. It works out how many draws the model expected across the exact fixtures a run was made of, and then — the part that changes the answer — how many runs that long ought to exist somewhere in a database this size. Out of half a million matches, a one-in-a-million run is not a miracle; it is a Tuesday. The same reading sits on each record run's own page, so a long run never gets to look rarer than it is.
History vs the market asks the blunter question: even if the pattern is real, was it ever worth money? It groups every priced match by how long that team had gone without drawing, and sets the rate they actually drew against what the bookmaker's closing price said the chance was, with the margin removed. The two agree. The price already knows what the run length says — including that teams on long runs draw less often, not more. There is no overdue edge left in a run length to collect. It needs a sign-in.
People chase draws with a doubling-up ladder whether or not a site tells them to, so this part of FDB4 does the arithmetic in the open rather than pretending nobody does it. It recommends nothing and predicts nothing. It shows what a plan costs.
The bank planner lays a Fibonacci stake ladder along a team's real fixture list and answers the only question that matters up front: how much money must be in hand, by which date, to still be betting at the twentieth match — and what you lose if the draw never comes. The numbers get uncomfortable fast, which is the point of showing them.
History vs the market then replays that plan over everything in the database: how many bets a line actually took, how often a given bank would have run out, and what every bank size would have returned. Every size loses, at roughly the bookmaker's margin, and a bigger bank only trades rarer disasters for bigger ones. Lines is the tracker for one person's own bets against that plan. All of these need a sign-in, and none of them is a tip sheet.
Everything above is computed from one archive of results, and the archive is browsable in its own right: results and fixtures by tournament, edition, team and date; tournaments, teams, league tables, head-to-head and Elo ratings.
Charts is the whole archive drawn rather than tabulated: the record runs laid on one calendar, draw rate by era for the competitions with the deepest history, and how far back our coverage actually reaches country by country. That last one is worth a look before you trust any figure on the site — the record is much thinner in 1950 than in 2020.
If you would rather compute your own answers, the REST API returns the same figures as the pages, and the MCP server exposes them to an AI agent directly. Both hand over data and caveats, never conclusions.
There is a red NOT BET ADVICE badge in the header of every page, and this page agrees with it rather than talking around it. FDB4 sells no picks, publishes no tips and has no opinion about your next match. Every number here describes matches that have already been played; nothing here forecasts one that has not.
Where a pattern turns out to be worthless, the site says so on the page that shows the pattern — that is what sections 3 and 4 above are. If you are looking for someone to tell you what to back, you are on the wrong site, and that is deliberate. The full disclaimer →