For a match to be counted, the results must be submitted for scoring. The winning team is responsible for submitting the results. We’ve attempted to streamline the process as much as possible. If it makes it more fun, you can imagine your in-universe Summoner grumbling about paperwork as you fill out the match submission form.
Submit Results Here
You’ll notice that we no longer require screenshots, which we hope will make things less cumbersome. Instead, we’re asking you to list the names of all participating Summoners. That makes duplicate identification easier, and more importantly, drastically reduces the need for scorers to manually look up individual Summoners’ ranks: we should be able to have the system automagically look up each Summoner on the master index. That also means that rank corrections automatically apply to prior matches.
Before I get into the details of scoring, here are some general rules. (Note: these were the rules as of Lines in the Sand. They have not yet been updated here. This version is left in place for reference.)
- We use the higher of Season 4 and Season 5 rank. If a Summoner was unranked in both Season 4 and Season 5, then and only then will we resort to their Season 3 rank. If they were also unranked in Season 3, they are listed as “unranked”.
- Because “unranked” Summoners have wildly varying skill levels, ranging from “I just hit level 30” to “I’ve been playing since beta but hate Ranked”, we reduce the value of a match for each unranked Summoner. (We compute the % of Summoners who are unranked in a given match, and reduce the value by that amount.) We’d like to develop a better rule, e.g., using LoLKing’s estimated MMR, so we can get rid of this interim measure.
- Our matchstaff use an “unranked = Gold” interim rule for Featured Matches. This rule will also be changing soon.
- The definitive source for S4 and S5 rank is Crazy MLC’s rank checker, which directly queries the Riot API. Sites like LoLKing also provide this information, but we can’t be sure of their methodology, and sometimes those other sites seem to get it wrong.
- A match is invalidated if a Summoner is playing against their own faction.
- Reduce a match value by a percentage equal to twice the percentage of Summoners in the match who are subs, i.e., playing for a faction other than their own. For example, if 2 Summoners are subs in a 5v5, the match value is reduced by 40%.
- Adjusts can reduce a match’s value to zero, though (of course) not below zero.
- Mostly for the sake of administrative simplicity, and also to avoid fuzzy questions about the timing of promotions, each Summoner is given one “effective rank” that applies for all matches in a given arc.
- Sometimes, it is necessary to calculate Balance of Power, e.g. for a lore reason, or for a tournament structure reason. We will update the BoP prior to making such determinations. Once those determinations are made, though, they are “locked in”, even if (e.g.) we later discover that we had the wrong rank for a Summoner. This is because of the horrific time travel headaches we get otherwise. In the same way, even if someone dug through the old files and discovered that Mirrorwater was improperly scored at one crucial point, so that Ionia should have won the Mirrorwater instead of the Freljord, we are not going to retcon the entire Factions storyline to accommodate this result. (Though we might write something into the story that would allow Ionians to justifiably accuse the Freljord of meddling.)
How Scoring Works
Note: I am currently conducting a survey of opinions on match-scoring. All are welcome to vote.
Factions matches are scored based on factors including:
- The type of match and number of Summoners
- The relative skill levels of the two teams
- The relative standings of the two factions
The winner gains this number of points, and the loser loses the same amount, making scoring a zero-sum system.
Base values (mode adjusts)
A 5v5 Summoner’s Rift match has a base value of 10 points. Other match types generally are worth fewer points.
Base value is calculated as follows:
- One point for every Summoner in the match, but with a 25% reduction if the match is being played at less than full capacity. (Because the removal of even one Summoner can significantly change the dynamics of the mode, making skill adjust less accurate.)
- Reduce the value for matches held on Fields of Justice other than Summoner’s Rift. This is mostly because our skill adjust system is not well-tuned for other Fields. The Twisted Treeline is worth the second most points, because its playstyle is fundamentally similar to a smaller Summoner’s Rift. The Howling Abyss is next, as it is essentially one big “teamfight simulator”. The Crystal Scar is worth the least points, because of two major issues: first, that Dominion has a very distinct skillset, and second, that Dominion matches are about half the length of other matches. I think the latter factor is a good reason for Dominion matches to be worth 50% points, and that value is further reduced due to the skill-adjust problem.
- Twisted Treeline: 80% points
- Howling Abyss: 50% points
- Crystal Scar: 30% points
- If the match is a Featured Match, increase the value to 150% of normal.
There are two main adjusts after the base value has been set:
- Skill Adjust. We check the rough skill level of each team’s Summoners, and adjust the match value accordingly. If the over-stacked team wins, it’ll gain fewer points, and its opponent will lose fewer points, than if it had been a fair fight.
- Faction Strength Adjust (or Rubber-banding). We also adjust the match value somewhat based on the relative standings between the two factions. For every point that separates the two factions’ scores going into the match, the value of the match is increased or decreased by 1%. This encourages “weaker” factions to fight “stronger” factions, and discourages stronger factions from always beating down on the same weaker factions. They can still do so, but an upset victory by the last place faction against the top-ranked faction will be worth bonus points.
The skill adjust is explained in more detail below.
Match values are adjusted based on the skill matchup between the two teams. After all, if a team of Noxian Plats beats up a team of Demacian Silvers, it would seem kind of unfair to give Noxus full points. Conversely, if those Demacian Silvers manage to beat a bunch of Noxian Plats, it seems like they should be getting some serious bonus points for overcoming that challenge.
Our specific goal is to adjust points based on rank such that, if both sides played at exactly their usual skill level, they could play a thousand matches and neither one would gain any significant net points. In other words, we want to compare the actual Factions match outcomes against a hypothetical “everyone playing solo queue with full rosters at their usual level” reality. The main factor we’re trying to “correct for” is base skill. Factors we are not correcting for, and in fact want to measure, include:
- The strength of each faction’s roster
- The teamwork among a faction’s Summoners
- Any greater passion or factional fervor than would be found in solo queue
- How well the faction’s Summoners have learned to use their roster
- How well the faction’s Summoners have learned to counter their adversary’s roster
- How high above their usual level Summoners rise during the match
If all of these things are “zero-impact”, then the rank adjust is meant to prevent any net points from moving around, over a large enough number of matches.
Note: There is a partial exception, in that we do apply a “rubber-band” adjust based on faction strength. It’s a relatively minor adjust, and it’s not based on any particular analytic objective. We just want to give weaker factions a little encouragement to tackle the top dogs every now and then. Unlike the skill adjust, it is not intended to hold factions at net-zero points, but merely to influence the rate of change.
Modifying the match value
If the system says that Demacia has a 70% chance of success based on skill imbalance alone, we want Demacia to take fewer points
You can also think of it with a dice game analogy. Double-or-nothing is a pretty fair betting rule in a coin-toss match. You put in one gold coin and then have a 50% chance of two gold coins and a 50% chance of zero gold coins. The expected payoff is one gold coin. If you played that game with one thousand gold coins, you’d probably come away with…one thousand gold coins. If the odds of winning drop to 25%, and it still costs one gold coin to play, you’re going to expect a payoff of four gold coins. This is essentially how the balancing works. The difference is the factors listed above.
To implement the system above, of course, we have to have some idea about odds of winning for various matchups. This was actually something of a difficult analytic problem. One idea was to perform a regression analysis; another was to create a sort of Elo system modeling each tier as a “player”. I later settled on the simpler idea of just looking at observed winrates. But before we could do anything at all, we needed more data.
Early in Factions history, we just arbitrarily assigned point values to the various ranks. (For example, a Silver might be worth 2 points, while a Gold might be worth 3.) We applied an arbitrary adjust based on this. We openly admitted that it was based on nothing more than supposition and guesswork. However, our feeling was that some adjust was better than no adjust. While far from accurate, the system did discourage “beatdowns”, and encouraged lower-ranked Summoners to get into the fight by assuaging their fears that they’d hurt their factions by playing.
Thousands of matches later, we have enough data to do some empirical analysis on how the different matchups work: for example, what the odds are for a team of two Golds and three Silvers (GGSSS) against a team of a Plat, three Silvers, and a Bronze (PSSSB). (As of this writing, the data indicates 61% odds of victory for the GGSSS team.) We have to use a few hacks to get a usable winrate out of this data, such as simplifying matchups by “cancelling out” mirrored Summoners: for example, GGSSS vs. PSSSB is actually processed as GG versus PB, with the three Silvers on each side being “cancelled out”. (You can check various matchups’ predicted winrates with this winrate calculator.) Also, if there’s not enough data in the system even after cancellation, the algorithm will try condensing Diamonds and Plats into a single “High” tier, and Silvers and Bronzes into a single “Low” tier, increasing the available number of datapoints.
There is still plenty of work to be done in terms of enhancing the analysis. One cool thing about this system, though, is that it’s inherently adaptive: every match played makes it smarter. (“But the best part…is that he’s learning.“) So when you send in match results, you’re not only making sure your faction gets due recognition for its victories: you’re also making Factions more fair in general.