the F41 [Archive]

[Chariot of Fire]

.....

[josko.ri]

I knew we'd lost a lot of points in Random League lately, and I guess this highlights exactly how much.

I see there are already many excuses how TOFU are F400 point victims by just participating in the Random League while it is for not true that they lost a lot of points there and it is for sure not the main reason why discrepancy in ranking between F400 and F41 is so huge. I will explain the real reason here, from mathematical point of view (roughly approximated):

1. When KORT (ranked 1393) lost vs TOFU (ranked 1357) by only 2 games 29-31 then KORT have lost 44 points.

Conclusion: losing by so small margin of only 2 games from clan who is 36 points lower ranked costed 44 points.

2. Let's consider TOFU's results in CL4 Phase 1. Majority of CL4 Phase 1 results were included into ranking around July 2012 so for this consideration I used July 1st 2012 ranking. TOFU had 1332 points, and average rank of all 7 opponents in CL4 was 1255. TOFU achieved total score of 39-45.

Conclusion: if from example (1) losing clan was ranked 36 points over winning clan, and lost by margin of 2 games in set whose weight was 61 was penalized by 44 points, then in example (2) clan who was ranked 78 points above average opponents' rank and lost by average margin of 6 games in challenge whose weight was 84, would for sure lose around 100 points from that challenge.

All 3 factors are unfavorable conditions in example (2) compared to example (1) (78v36 point difference, 6v2 losing margin and 84v61 total challenge weight) so that multiplies total points lost from the challenge. The only favorable condition for example (2) compared to example (1) is decay which happened in timeframe of 1 year, in comparing with no decay in example (1).

3. For Random League consideration, I will use TOFU's rank as 1375, which is median between their rank before and after KORT challenge, so half of RL results were included with old rank and half with new rank. For average rank of their opponents I will use May 1st F400 because it is median timeframe. Average rank of the 12 opponent clans that they completed their sets by using May 1st F400 ranking is 1006. From the 12 completed sets (144 games), TOFU's score is 93-51.

For comparison and concluding how many points this can cost TOFU or get to TOFU, I was searching for another match of similar ranking opponents and similar final result. I found IA (1263) vs LEG (922) [341 point difference] which ended 28-13[68,3%] from July 2013 ranking as a fair comparison to TOFU (1375) vs average RL opponents (1006) [369 point difference] which ended 93-51[64,6%]. Point difference between IA and LEG was little lower, but total score was little higher so ranking points awarded should be more or less equal. IA neither won neither lost points for winning that challenge because they had 1263 points both in July 15th and August 1st 2013, and the result was included in ranking at July 29th.

Conclusion: If IA which is very similar case did not lost nor get any points in war of 41 games, then also TOFU for similar winning and ranking difference conditions should stay more on less on the same number of points like they would have if they did not play RL. Weight of the challenge in this consideration can be neglected because IA had 1263 from their 380 games in last 2 years, and to that was added win of 28v13 which obviously had point value of around 1263 and weight of 44 games, so if weight was higher it would still be the same in averaging with their previous score.

Final conclusion and proof that Doc_Brown's accusing participation in RL for point lost is false:
From this brief mathematical analysis and comparison of similar results and rankings and points awarded for wins/losses, it is obvious that bad results in CL4 vs similar ranked opponents costed TOFU around 100 points while their results from RL3 approximately neither give neither took out points from them, so my initial opinion that the quoted post is lie is proven.

That was final conclusion about TOFU's results only, and here is consideration about KORT's results from small sets which had significant weight:
In CL4 Phase 2, KORT achieved score of 46-38 vs clans which were ranked in average around 100 points below. In CL4 Phase 3, KORT achieved score of 32-22 vs clans ranked around 100 points below. Those 2 sets are 2 sets with the highest weight from small challenges in last 2 years of KORT's activity, and those 2 sets combined for sure were awarded KORT by around 50 points. I did not combined CL4 phase 1 and CL5 phase 1 because on those competitions both clans achieved more or less similar results vs more or less similar ranked clans.

Final explanation for discrepancy between KORT's and TOFU's ranks in F400 and F41:
By comparing 2 major sets which were played by TOFU (CL4 P2 and RL3) and KORT (CL4 P2 and CL4 P3) and which are lower than 41 game sets, it is approximated that TOFU lost around 100 points in CL4 P2 and zero points in RL3, while KORT gained around 50 points in CL4 P2 and P3. That difference of around 150 points is approximately similar like difference between their F400 and F41 ranking which is 128 points. But source of where points were lost by TOFU is completely different than the source which has been been pointed by Doc_Brown, it is CL4 where KORT gained a lot of points while TOFU lost a lot of points, and not RL3 where TOFU neither lost neither won points.

My calculation is brief and approximately, for exact numbers IcePack can provide the real analysis from the Algorithm if he wish and has time, but I guarantee that real points gain/loss is not far away from my estimation made here.

[Doc_Brown]

Josko,
Hopefully this doesn't go negative as quickly as some of the other conversations, but let's give it a shot. This is an interesting analysis, and I think you have some valid points. However, there are some mathematical flaws. First, the CL4 phase 2 batches are quite a bit older, and if I understand the ranking method correctly, they don't count as much as recent wins/losses (I haven't looked at the algorithms in much detail so I may be wrong on this point). On the other hand, your goal is to compare TOFU to KORT in relative terms, so this isn't as big a deal.

A more serious math issue is in terms of strength of opponents. Simply using the average rating of opponents is going to get you in trouble. As an example, assume you are a 2000 point player and play two games, one against a 1000 point player and one against a 3000 point player. If you beat the higher ranked and lose to the lower ranked player, you'll end up with you'll end up with a net loss of 10 points, or if you lose to the higher ranked and beat the lower ranked player, you'll have a net loss of 3 points. However, you went 50% against players with a mean ranking the same as yours, so your analysis would have suggested a net zero point change. In stock market terms, this is known as "draw down." Actual performance over an entire group will be less than expected based on the average of the group.

All that said, it looks like you're right that the biggest negative factor in our score is the CL4 results. But I think that's a good thing. That means the worst scores are some of the oldest, and those will be some of the first to fall off the back of our rankings. So TOFU's overall score should climb in the near term. Once those fall off though, I think the RL games will be the worst results, and they're certainly the worst results we've had since I joined the clan in January.

The one other point I wanted to address was this comment:

... so my initial opinion that the quoted post is lie is proven.

The primary definition of a lie is "an assertion of something known or believed by the speaker to be untrue with intent to deceive," whereas a mistake is "a wrong action or statement proceeding from faulty judgment, inadequate knowledge, or inattention." You have logically supported an assertion that my post was a mistake. To call it a lie is to impugn my character and suggests that I am purposely ignoring facts rather than simply being unaware of them. I'd again point out that I joined the clan in January, and I have, at best, a cursory knowledge of the results TOFU has received prior to 2013. Since you are claiming that results from 2012 are the biggest draw on our rankings, and not the RL results from 2013 that I am aware of (though I think you would agree that the RL results are much more negative - or, less positive if you prefer - than the CC3, CC4, and CL5 results we have received in 2013), this would fit in the "inadequate knowledge" category.

I'll give you the benefit of the doubt and assume that your statement was a result of inadequate knowledge of English and was therefore a mistake rather than a lie.

[Chariot of Fire]

Don't waste your breath mate, not worth it.

[hotfire]

awesome work icepack...im amped u did this

[Chariot of Fire]

Just to answer Josko's Pt.1

1. When KORT (ranked 1393) lost vs TOFU (ranked 1357) by only 2 games 29-31 then KORT have lost 44 points.

Conclusion: losing by so small margin of only 2 games from clan who is 36 points lower ranked costed 44 points.

Your conclusion is wrong. During the period 1-15 May 2013 for which IcePack did the F400 there were other considerations to be made.

1. KORT had results other than the 29-31 loss to TOFU, so you cannot say the TOFU result cost 44pts
2. KORT will have lost points due to ageing of the May 6th 2011 win over THOTA
3. TOFU during 1-15th May 2013 recorded wins of 31-29, 14-2, 9-3, 12-4, 7-5 & 10-2....yet our points increase was a lot less than the 44 that KORT lost, so it is quite clear the singular TOFU-KORT result did not cost your clan 44pts.

So get your facts straight Einstein before you go accusing people of being a liar.

[josko.ri]

I plan to respond just to replies regrading math and F400 algorithm, not to other comments.

A more serious math issue is in terms of strength of opponents. Simply using the average rating of opponents is going to get you in trouble. As an example, assume you are a 2000 point player and play two games, one against a 1000 point player and one against a 3000 point player. If you beat the higher ranked and lose to the lower ranked player, you'll end up with you'll end up with a net loss of 10 points, or if you lose to the higher ranked and beat the lower ranked player, you'll have a net loss of 3 points. However, you went 50% against players with a mean ranking the same as yours, so your analysis would have suggested a net zero point change. In stock market terms, this is known as "draw down." Actual performance over an entire group will be less than expected based on the average of the group.

So from your example, you will get higher benefit (-3 points) if you lose to higher ranked than if you lose to lower ranked (-10 points). So, let's see in average who TOFU lost (and draw) vs, were them lower or higher ranked than TOFU? Lost sets were vs LOW (1099) and OSA (1139) while draw was vs TFFS (994) while average opponents' rank was 1006. So, following your logic, TOFU even got a profit in comparing with averaging opponent rank, because they lost points vs opponents who are the same like average (TFFS) or much stronger then average (OSA, LOW) which means they achieved wins vs clans lower ranked than average, which is with your logical explanation scenario which would give to TOFU more points than what they would get in comparing with average clans' rank. As conclusion, if my previous consideration lead to result that no points were gained nor lost in RL3, then with your new input conclusion is that TOFU even gained some points from RL3.

Just to answer Josko's Pt.1

1. When KORT (ranked 1393) lost vs TOFU (ranked 1357) by only 2 games 29-31 then KORT have lost 44 points.

Conclusion: losing by so small margin of only 2 games from clan who is 36 points lower ranked costed 44 points.

Your conclusion is wrong. During the period 1-15 May 2013 for which IcePack did the F400 there were other considerations to be made.

1. KORT had results other than the 29-31 loss to TOFU, so you cannot say the TOFU result cost 44pts
2. KORT will have lost points due to ageing of the May 6th 2011 win over THOTA
3. TOFU during 1-15th May 2013 recorded wins of 31-29, 14-2, 9-3, 12-4, 7-5 & 10-2....yet our points increase was a lot less than the 44 that KORT lost, so it is quite clear the singular TOFU-KORT result did not cost your clan 44pts.

So get your facts straight Einstein before you go accusing people of being a liar.

In the same ranking update, TOFU got 36 points and KORT lost 44, so point value gained/lost from that war is somewhere around that numbers, in my consideration it is irrelevant is it 36,37,40 or 44, those numbers are all in the same range anyway.

As for your "advice" about getting facts straight before accusing people to be liar, the KORT-THOTA war were included in ranking with date KORT z-THOTA 33 27 04/19/11, which means that on May 1st ranking that result was already aged and does not have influence on May15th ranking. I think I do not see the first time that you are pointing fingers at me for telling untrue facts, while actually you are the one who speaks untrue facts my friend

Looking at only KORT's results from May 1st to May 15th, there were the match vs TOFU, and 13v3 win vs DB, which is value around 1350 ranking, so that one result did not have big influence to our rankings. That said, KORT's net loss of points is more similar to real net gain/loss of points in that war than TOFU's net points won, because TOFU had 5 other mini wars in addition to the big war while KORT had only 1 mini war included together.

As a conclusion, no matter how much you try to go around facts, the net point loss was around 44, because weight of that challenge was 61 while weight of other included war was only 16 and result in the other war was not out of average for given clan ranks of KORT and DB.

Anyway, of the whole my initial post the most analysis was written in point 3, which for sure is the most important point, so here I will quote the most important conclusion which considers RL3 results only:

For comparison and concluding how many points this can cost TOFU or get to TOFU, I was searching for another match of similar ranking opponents and similar final result. I found IA (1263) vs LEG (922) [341 point difference] which ended 28-13[68,3%] from July 2013 ranking as a fair comparison to TOFU (1375) vs average RL opponents (1006) [369 point difference] which ended 93-51[64,6%]. Point difference between IA and LEG was little lower, but total score was little higher so ranking points awarded should be more or less equal. IA neither won neither lost points for winning that challenge because they had 1263 points both in July 15th and August 1st 2013, and the result was included in ranking at July 29th.

Conclusion: If IA which is very similar case did not lost nor get any points in war of 41 games, then also TOFU for similar winning and ranking difference conditions should stay more on less on the same number of points like they would have if they did not play RL. Weight of the challenge in this consideration can be neglected because IA had 1263 from their 380 games in last 2 years, and to that was added win of 28v13 which obviously had point value of around 1263 and weight of 44 games, so if weight was higher it would still be the same in averaging with their previous score.

[Armandolas]

Who cares where someone lost points?..To.FU is Number 1 in both rankings....But what matters really is to congrat Icepack fot putting an amazing work and effort that makes this rankings possible

[Doc_Brown]

I plan to respond just to replies regrading math and F400 algorithm, not to other comments.

A more serious math issue is in terms of strength of opponents. Simply using the average rating of opponents is going to get you in trouble. As an example, assume you are a 2000 point player and play two games, one against a 1000 point player and one against a 3000 point player. If you beat the higher ranked and lose to the lower ranked player, you'll end up with you'll end up with a net loss of 10 points, or if you lose to the higher ranked and beat the lower ranked player, you'll have a net loss of 3 points. However, you went 50% against players with a mean ranking the same as yours, so your analysis would have suggested a net zero point change. In stock market terms, this is known as "draw down." Actual performance over an entire group will be less than expected based on the average of the group.

So from your example, you will get higher benefit (-3 points) if you lose to higher ranked than if you lose to lower ranked (-10 points).

Actually, you missed the point. In the example I gave there are three scenarios:
1) Looking at the average, a player went 50% against teams with an average ranking identical to what he currently has. Theoretical result: net change of points is 0. (This scenario is the estimated results you're offering.)
2) Exact calculation assuming the player loses to the low rank and beats the higher rank: Actual loss is 10 points (worse than theoretical estimate).
3) Exact calculation assuming the player beats the low rank and loses to the higher rank: Actual loss is 3 points (worse than theoretical estimate).

Your estimates say that TOFU should have gone effectively net zero from RL games. What I'm showing with a comparable example is that because of draw down, the actual point gains will, of mathematical necessity, be lower than the theoretical estimate. Some scenarios are better than others, but all scenarios are worse than predictions based on the average.

Anyway, it's an interesting discussion, and I appreciate all the time you spent digging through TOFU's records. As I said, I think your estimates are at least partially valid, though I think one would have to look at it on a war-by-war basis to determine the true effect of the RL matches on TOFU's record. Although given the way the most recent batches have gone, if we didn't have a negative result before, we probably will soon.

And I'll chime in with Armandolas in thanking Icepack for all of his work. More than anything, I just find the results fascinating, particularly given the differences between F400 and F41. I think this discussions started with my surprise at how much higher TOFU's score was in the F41, and I stand by that. If the cause of it is CL4 batch 2 rather than RL3 batches, that's fine. It really doesn't matter a lick. In fact, I'm hoping Josko is right because it means our F400 score will gradually drift upwards towards the F41 score as the old CL4 batches drop off, rather than a drift in the other direction that would result if the RL games were the source of the difference.

[IcePack]

So...I stayed home with a migraine all day. To be honest, most of the comments from kort/tofu were just plain tl:dr.
I will look at them in more detail another day.
But, I was asked by someone to go back and discuss what really went on between the two clans on that fateful day in May.

So, I'll share a QUICK review on that war. (please dont make me regret it) I'm 99% sure I covered all my bases.

TOFU vs KORT completed 5.2.13; the day before (5.1.13):
KORT @ 1388
TOFU @ 1348

The day results entered:
TOFU @ 1383
KORT @ 1354

No other entries were entered from 5.2.13 that would have affected either clan. Decay looks like it was 1 pt on kort (without the war).

Sorry its so brief. Good luck with the rest of the argument.

[hotfire]

lol :^)

[agentcom]

I was thinking josko's estimate of the lost points sounded a bit high ... but yeah tl;dr past that point

[IcePack]

Updated

[Gilligan]

Christ almighty, number 6

[Timminz]

Go go rocket-bots!

[IcePack]

Christ almighty, number 6

Almost done with F400....

[maasman]

No LoW yet, eh?

[IcePack]

No LoW yet, eh?

Nope, LOW has 90 weight in the F41 and needs 125. Though, they definitely are not as strong on F41 as other areas of rankings...

[maasman]

No LoW yet, eh?

Nope, LOW has 90 weight in the F41 and needs 125. Though, they definitely are not as strong on F41 as other areas of rankings...

Completely understandable seeing as we haven't played many full challenges lately and have been doing well in other areas. I imagine after the ID challenge we'll be included.

[Bones2484]

Christ almighty, number 6

Wow. We might just break the top 20 in the coaches poll now.

[IcePack]

Edit: (error)
MB listed twice. Bottom #39 is VVV.

[IcePack]

Update here planned for Oct 1st.

[IcePack]

Updated, very few wars completed so kinda truncated the info.

[IcePack]

Archive

[Chariot of Fire]

I've given up trying to figure out how this and the F400 are calculated. With a recent 46-15 demolition of Dynasty we still manage to lose a shedload of points.