An alt-3 ranking takes proper account of each team’s schedule strength — that is, whether they have played (on average, to date) relatively ‘easy’ or ‘hard’ opponents. The easy/hard assessment is made in a mathematically coherent way, and is based only on the current season’s match results. Home advantage (or, perhaps, disadvantage) is automatically taken into account, by recognising that some teams do better when playing at home than when playing away (or, perhaps, better away than at home).
The alt-3 table displays two new columns, labelled ePld and Rate. We describe here, just briefly, what each of those columns refers to.
For (quite a bit!) more detail, please see the companion page to this one: The mathematical method explained.
The two columns are as follows:
ePld, the assessed effective number of matches played by each team.
This reflects how tough a schedule each team has faced, since the start of the season. An ‘average’ schedule results in ePld being the same as the actual number of matches played to date. But if a team has faced a particularly tough set of matches so far — i.e., matches against a stronger than average set of opponents — then that team has in effect used up fewer of their realistic chances to accumulate league points than their actual number of matches played would indicate. The team’s value of ePld quantifies this: the difference between the actual and effective matches played
schedule strength = Pld − ePld
tells us how many schedule-strength-related ‘games in hand’ each such team has.
Similarly for teams whose fixtures to date have been easier than average: those teams’ ePld values will be larger than the actual number of matches played.
In the alt-3 table, teams that currently have 1.0 or more schedule-strength-related ‘games in hand’ are highlighted in light green: such teams are actually doing better than their accumulated league points would seem to indicate. Teams whose ePld number exceeds their actual number of matches played are highlighted in light red: those are the teams doing appreciably less well than their accumulated league points would suggest.
Where schedule-strength imbalance is so strong that ePld and ePld differ by more than 2.0, the entry in the ePld column is shown in bold type.
Rate, the league points per effective match played, for each team.
This is the column that determines the ordering of teams in the alt-3 table.
The Rate — or league points per effective match played — is (normally, i.e., for teams that have not had any points deducted by the league) just the ratio of two other columns in the table:
Rate = Pts ÷ ePld.
Where any team has had points deducted by the league, the Rate formula is generalized to
Rate = [(Pts + deduction) ÷ ePld] - [deduction ÷ seasonLength]
where seasonLength is the total number of matches that each team will play in the full season (for example, in a 20-team league the seasonLength will be 38). In other words, the normal alt-3 Rate is adjusted downwards by the appropriate per-match points deduction.
Where a team’s fixture schedule to date is unusually tough, or unusually easy, the alt-3 Rate will be substantially more informative than the simple tally of league points accumulated, as an indicator of current standing in the league.
It really must be emphasised that an alt-3 table is not a prediction. The effective matches played, ePLd, and the consequent points-per-effective-match Rate, are designed specifically to reflect fairly the relative performance of teams since the start of the current season — that is, to represent faithfully what has happened, rather than to predict what will happen in the future.
A good predictive model would, instead, most likely need to take into account much more information than simply the match results seen already within the current season.
It would be of course be mathematically possible to turn the computed Rate values into numbers that correspond directly to end-of-season points totals (for example, in the English Premier League, we could just multiply each team’s computed Rate by 38 to do that). But such a ‘prediction’ would almost certainly be inferior to other possible methods — methods that are not constrained to use all of (and only) the match-result data from the current season, for example.
Just to reiterate again, because it cannot be said too often:
The alt-3 table is not a prediction!