Ok FT Knockout currently sets division matchups based largely on a the number of knockout points someone has, and a random element.

I've heard a number of complaints that this doesn't create very good matchups. For those who believe this, what should be the primary ordering variable? Level? Number of badges? Something else?

Not badges - I earned most of mine by playing quizzes and writing them to earn my points, not by playing timed games. I don't mind the current setup for myself - once I had risen to my level of incompetence through a bunch of relatively easy wins, I now sometimes win the first match, sometimes not, and reckon I get a weekly win at least two or three times a year. It's enough to keep me in the game.

I don't think there's any particular variable mix that would make the Knockout matchups any better. Now someone with 500 points and a weekly win % of 33% may get matched up often with someone that has 500 points and a weekly win % of 5%. The 33% player would likely be favored but as we see from the Fantasy KO underdogs win quite often. Sorting on Knockout points per week played would put people in similar tournaments week over week as well I'd imagine. I would honestly be in favor of having a larger random variable (at least 50% of the equation) for tournament placement with maybe some type of lower bound (i.e. someone with 500 points can't be placed below Division 40). I seem to be in Division 1 75% of the time.

Could it be multiple variables? Like a combination of average score in knockout games, weekly win percentage and total wins (plus some random variable just to make sure the brackets aren't almost identical most weeks particularly near the top.)

I don't see any problem with the match-ups as they are.

Hmmmmm for some reason I was under the impression that large numbers of people were unhappy with the bracket selections in this game, but it sounds like not.

I'm only unhappy with the bungee cord that keeps me tied to the upper groups as soon as I set foot in the game!

Probably just me then. I guess I'll just have to stop playing it for a couple of years so that I can drop down somewhere nearer to players of my own ability!

Let's give people a chance to see this thread - I know I've heard from some people who had issues, but I'm not entirely sure what they were, as I don't play the game.

I don't think you're the only one. I know I've considered dropping Knock Out from my list. The single digit brackets are matching the same people up very often, it seems. It doesn't take too many first and second round losses to wonder what the point is.

How about using a (total KO points/KO wins) as a sort? Newer KO players who are very successful would get a chance to play at a higher level and more seasoned players who have moved up based on longevity more than wins in the last few years would be back in more competitive brackets?

I don't consider myself among the group of people who are necessarily unhappy with the way KO Tournaments are determined. Having said that, it seems to me that the matches should take into account not only the total number of KO points, but also the ratio between total number of weekly victories and total number of KO points. Also factored in should be time since last weekly victory, if that already isn't being done.



On the FT Knockout FAQs page it states: "...while players with 1,000,000 points and 5 years of experience play against similar players." Yet I gather that isn't being factored in explicitly? So perhaps it would be a good idea to add this to the mix.

I like the look of this suggestion.

I think the main problem with the current metric is that you go up until you reach your proper strength level, then you get a few lucky wins and now you're playing above your strength and can't advance at all until the player base around you have all scored a lot of points to pass you again. In that time, you still score the occasional KO point and thus this phase drags out even more.

Adding in just a total of wins would make this a bit shorter but not really that much.

How about a metric that factors in relatively recent wins as a multiplier to KO points instead?

If you play a round and win, your multiplier goes up by 0.1.
If you play, but lose (not due a no-show but in an actual game), it goes down by 0.01.
Not playing or losing via no show does not change the multiplier at all.

Obviously there need to be an upper and lower cap for these, say 2.0 and 0.5.

So, as an example, I have 350 KO points and win my bracket. That gives me 5 points and also I take a 1.1 multiplier, so the following week, I would be paired as if I had 355 x 1.10 KO points, which is 390.5.

Then I next take a streak of beatings, with 24 weeks of no wins. In that time, I pick up another 17 points from occasional first and second round wins, which puts me to 372 base points. My multiplier is now 0.86 (.24 less than the 1.10 I had), so for pairing purposes, I now have 319.92 and probably a much better chance to win a round again soon.

With this metric, you'd go up against harder opponents more quickly if you win a lot but you also come back to more comfortable waters in a reasonable time if you don't.

I have a letter in my inbox from a player who has been playing Knockout for many years, and is often matched with players who have roughly the same number of points as he has, but who have double or more the number of wins. He feels way outclassed.


There was also a request to not have team members matched against each other, if this is something that can be easily done - a lot of people dislike knocking out a teamie.

I think the team aspect would be very difficult not so much from coding, but from balancing. A very strong team such as the Mossbacks would have the chance to dominate all the top divisions if only one member could be in each bracket.

If at all, it would probably have to be such that the tournament assignment is drawn randomly and only then any care given about teammates (i.e. if two members from the same team were in one bracket, they could only meet in the final, three or four at the earliest in the semis...)

My proposed algorithm would address your player's concern - having fewer wins, they'd have a lower multiplier than the ones who win regularly and thus not see those players again.

(By the way - the numbers might sound complex, but this is nothing the player would worry about or even see. They'd just be an internal number affecting who gets paired with whom)

Personally I don't feel the need to change the KO selection, but go ahead if there are too many complaints.

A side remark on the suggestion of using the number of wins divided by the number of plays: for people who play their very first week, this can mean a great distortion. Imagine a new player who wins all four matches on his first week: this player is bound to qualify for the top level, as his/her win percentage equals 100%.

I read Wes' post #1238833 above. My concern with that approach (further elaborated upon in the last sentence of post #1238848) is that, if implemented, we would soon see a different type of complaint:

Why am I always in a tournament with people with a whole lot more Knockout points than me? Even if they had seen Wes' posts (unlikely), given that the pairings would be determined by internal numbers not visible to anyone except administrators, there would be plenty of opportunity for such grousing. And even if the number of Knockout points was eliminated from viewing altogether in the tournament lists, then we would have:

Why am I always in a tournament with people at much higher level (or some other metric they could look up) than me?

I don't think there's anything that can be done to eliminate potential griping. And the thing about the Peter Principle is it gets you eventually no matter what metric gets used. If you set up a system that (even if totally unintended) rewards people for losing by making matchups more favorable thereafter, it is likely a few players will begin to underplay intentionally, something that would be next-to-impossible to actually prove. That is especially true when you have some players who are more interested in winning *tournaments* than winning individual Knockout *games*.

The people I have the most empathy for are the ones who have been playing for so long they are almost always in the top five tournaments, where there is always at least one superstar around who will make it almost impossible for them to win a tournament. Sure they can win *games*, but that won't help you upgrade your Diamond Knockout badge, if that is your goal.

Actually, underplaying would not gain you much in my setup. I only increase the multiplier when you win your tournament - if you lose (whether in round 1 or in the finals), your multiplier goes down. Thus, you can try your best every time you play.

The truth about KO is that in a bracket of 16, 93.3% of players will NOT win their round every week. If you play every week and all pairings are perfect (all players in the pairings are of equal strength), you will win the bracket on average only a little more often than 3 times a year. Things can quickly get very frustrating when you can't hit this average - not winning the bracket even once for 8 months or a whole year is not out of the question if you get overrated by even a small amount - if your per-match win rate only goes down from 0.5 to 0.4, your bracket win rate drops to 1 in 39 weeks (9 months). If your match win rate becomes 0.35, you're already looking at 66 weeks (15 months) between championships.

If we don't want to make major changes to KO timing (running more tournaments would obviously give players more chances), the pairing algorithm should probably take into account recent bracket win rates in some way so that players who exceed the already relatively long average time between wins get a better chance and stay motivated.

As an addendum, here's a little table on how the per match win chance translates into bracket win frequency:

75% - 17 wins per year
70% - 13 wins per year
65% - 10 wins per year
60% - 7 wins per year
55% - 5 wins per year
52.5% - 4 wins per year
50% - 3.25 wins per year
47.5% - 2.5 wins per year
45% - 2 wins per year
40% - 1.3 wins per year (9 months between wins)
35% - 1 year 3 months between wins
30% - 2 years 3 months between wins
25% - 4 years 6 months between wins

The post before his last one by wesleycrusher really sums up the way I feel about Knockout these days when he mentions motivation. There is next to no motivation to play it and my no show rate is starting to creep up because at this stage it is getting to be quite predictable. If I win the first game I definitely lose the second one. It’s frustrating too because as many who have read my comments might remember I am not one to shirk a challenge. I don’t mind a situation where it’s really tough to win, but this has started to test me to the point where I had already decided to stop playing it for a substantial length of time to try to reset my level. That’s a shame because it’s one of the best games on the site.

Up until three years ago I had been getting one weekly win per year for the three or four years prior to that. In these last three years I’ve had just two weekly wins and my current barren run stretches back to April 2018.

Now Brian might be right. Maybe I won’t win under a different selection criteria either. I’m willing to give it a go though. Having the chance to feel competitive would be a start! It’s never going to be my best game, I know that, maybe my win rate is correct for me. Not everyone can win! There’s gotta be a loser, so maybe I’m just a loser for Knockout! It just feels like I’m penalised for having played it in a mediocre way for too long.

I don’t want to be the one to advocate this strongly either if most people are happy with it the way it is. It’s still a very good game and shouldn’t be amended just because one (or a few) person has become a little tired of it. It can be my choice or others choice whether to play it or not.

I'm getting notes about it from people who are not comfortable posting here. Feeling penalized for having played in a mediocre way for too long is the common feeling - anything that can be done to help that would be an improvement.

Thanks for that stat, Wes. It puts it in perspective for me.

I always play KO, every week and think I've been a no-show maybe twice since it began. I seldom win my bracket, but continue playing because I like the challenge and also because it gives me satisfaction to win any match-ups at all, even the first one. I once went over a year, I think it was very close to two years without winning my bracket, but I always lived in hope!

After seeing the analysis and numbers (they were as surprising to me when I calculated them as they may be for many when reading them), I start to feel that a change to the game itself might be in order. But before I get to that, I want to put some background as to why this seems to have become an issue now far more than back then - the game has been an unchanged part of the site for many years, right?

Turns out that is not the case. When I joined in 2008, the game was just about KO points. You got a badge for 50 and another for 200 and that was it. Winning the weekly bracket was a little bonus (you got that 5th point), but not really important. Players thus got between one and four chances each week to make progress - and each of these chances was somewhere between 30% and 70% (the chance to win one match).

Nowadays, we have the Emerald / Ruby / Diamond KO series with their upgrades. You get one shot a week at this series, with a chance anywhere between 0.8% and 24% (using the same 30% and 70% match win percentages). And this chance becomes really low once you have to play beyond your strength.

This yields an unwanted situation: You have a low chance at succeeding at each individual attempt AND you have to wait a long time to get another try.

Fixing the assignment algorithm can help the "low chance" part to some extent, but even with perfect fairness, we're still looking only at the 1/16 chance of a coin toss (and it would be a boring game if the pairings neutralized every skill advantage).

Thus - what if we change the frequency of the game?

My suggestion is as follows: Do away with the Monday signups. Instead, you can start a bracket every day UNLESS you are still active in a game by virtue of having advanced. If you lose your round, you immediately restart in round 1 of a fresh bracket of 16 on the next day. If you keep winning, you play the next day in round 2, 3 or 4 of your ongoing bracket. Either way, you play exactly one 20-question set per day for as long as you remain subscribed (this would include Saturdays, Sundays and Mondays).

This means those who win often get close to twice the chances. For those who tend to lose a lot however, you get up to seven times the chances. And an average but daily player (with a 50% chance to win each round) would see about 180 brackets - and close to 12 bracket wins - a year.

Fantasy KO would still be a weekly game - it would simply take tournaments that start on a Tuesday and randomly pick seven of those.

(Note, I haven't checked this with Terry, but I think this could be doable, code-wise).

I don’t blame them! I’m just daft enough to not care about making a fool of myself repeatedly on the forums and on the chat boards!

I must say that I don’t like the idea of rolling tournaments that you suggest Wesley, KO is beautiful in its simplicity. Such an elegantly constructed game, I like the structured nature of it.

It seems like there would be another advantage to this. On the 2nd day, it would be only players who lost the first day so the really top people at any level would not be in this round. Although this could average out over a few weeks.