Data Collection: Raid Matching

lol no deal, I’m kinda short on gloves too :laughing:

1 Like

Thanks but I wasn’t born yesterday :joy::joy::joy::joy::rofl:

Ask @Shohoku79

Are you logging all of your matches, including those you decline in favor of a reroll?

I haven’t rerolled yet, those are the only matches I’ve had so far. (But I intend to log rerolls too.)

EDIT: FYI, I don’t tend to reroll a ton when I’m not filling a chest, and even then I don’t always bother. Unless I’m really in need of resources, I raid outside of chest filling just for fun and practice for War. And I routinely get matched with teams with 400-800+ tp more than I’m attacking with.

Result Key:
R: Reroll
W: Win
L: Loss

Note: I haven’t included rematches or revenges here just the first outcome of the pairings given by the Find Opponent button.

My Cups Their Cups Their TP Result
1950 2214 3661 W
2000 2118 3450 W
2013 2205 3777 R
2013 2292 3808 L
2002 1765 3673 R
2002 1862 3165 R
2002 2014 3560 R
2002 2189 3886 R
2002 2073 3408 L
1978 2203 3574 R
1978 1944 3575 R
1978 2016 3487 R
1978 1842 3177 W
1997 1898 3376 W
2001 1842 3271 R
2001 2092 3691 R
2001 2201 3669 R
2001 1997 3421 R
2001 2184 3637 R
2001 2160 3534 R
2001 2107 3514 R
2001 2041 3431 R
2001 2081 3441 W
1832 1914 3583 R
1832 2083 3453 R
1832 2106 3421 R
1832 2041 3405 R
1832 2083 3544 R
1832 1555 3038 R
1832 2058 3454 W
1880 2041 3296 W
1923 1990 3692 R
1923 2129 3687 R
1923 1949 3391 W
1956 2218 3797 R
1956 2061 3303 L
1918 2078 3513 R
1918 2182 3825 L
1892 1619 3179 W
1903 2155 3571 R
1903 1909 3449 R
1903 1682 3123 R
1903 2159 3843 R
1903 2010 3450 W
1942 2031 3567 R
1942 1917 3398 L
1931 1757 3161 R
1931 2055 3474 L
1940 2184 3571 R
1940 2217 3922 R
1940 1983 3310 W
1974 1993 3245 W
1963 2201 3852 R
1963 2009 3471 R
1963 2203 3511 L
2000 1934 3266 L
1934 1971 3525 R
1934 2067 3413 R
1934 1897 2124 R
1934 2198 3452 R
1934 2171 3806 R
1934 2043 3487 W
1974 1814 3421 R
1974 1766 3431 L
1948 1679 2985 R
1948 1897 3569 L
1945 2034 3438 R
1945 2187 3426 L
1871 1886 3315 W
1903 1932 3184 R
1903 1939 3377 W
1937 2069 3628 R
1937 2001 3479 L
1912 1948 3313 R
1912 2017 3415 R
1912 2145 3549 R
1912 2155 3453 R
1912 1790 3475 R
1912 1942 3277 W

8 posts were merged into an existing topic: [Primer] What do forum “Likes” do? Or use your “Likes”

Thank you. This data has been appended to my dataset. If/when you have more to add please do so as a separate post.

1 Like

Sounds good. I have not added your data to mine yet, so feel free to keep adding your data to your post above.

Will do, thanks! How much is reasonable to log? I was thinking maybe 100 matches?

Definitely when :slight_smile:

Will do. Thanks for coming up with the idea to do this, and for running the project! And for the interesting graphs so far.

I’d love to get as much as you are willing to do, particularly because you are collecting data from a trophy range that I likely can’t get to, at least not for a while as I am approaching the Level 70 Wall.

I may have to hit you up for some help on some of the stats. Your command of math seems significantly better than my own, or at least far less atrophied.

I’m happy to help in whatever way I can. Just let me know what I can do.

The main thing I’d like to do at some point is actually put some confidence intervals on the Trop v. TP regression curve. Once upon I time I knew how to do that, but that was a long time ago.

Ok, I’ll do what I can for data collection.

For what it’s worth, my defense team is maxed 4*, no 5*, and not super awesome ones for defense at that. I’ve prioritized titans and PVE over defense in my leveling choices.

And as you can probably tell from the attack team TPs I’m listing, I’m usually attacking with a mix of maxed 4*, unmaxed 4*, and occasionally a partially leveled 5*.

So the 4-70 wall on 4* shouldn’t keep you out of a similar trophy range, at least not intrinsically.

No worries. Here’s how to do that.

In order to find the 95% Confidence Interval (CI) on a simple linear regression slope, m, we calculate:

  • CI = m ± t(.0025, # samples – 2) * sqrt(MSE/SSX)

Where:

  • t(.025, # samples – 2) is the critical t value using a two-tailed t distribution at the 95% point, with (samples-2) degrees of freedom
  • MSE is the mean squared error on the regression
  • SSX is the sum of squares of the x values

MSE = sum((y_samples - y_mean)^2) / (samples-2)

SSX = sum((x_samples - x_mean)^2)

In Excel, the critical t value can be generated with:

TINV(a,df)

Where a is the confidence threshold, and df is the degrees of freedom. As the number of samples gets large, this will converge toward a Gaussian distribution’s critical value. You’re pretty much there with 700+ samples, but using the t distribution is just good practice.

1 Like

FYI in case you’ve been looking at my data as I continue logging, I still haven’t rerolled even once since I started logging. But I will log any rerolls.

First set of data - consecutive attempts. Headers as you suggested attacker cups, defender cups, defender TS

1481 1538 3280
1481 1710 2895
1529 1712 3429
1529 1635 3420
1529 1784 3341
1529 1722 3313
1529 1623 3043
1529 1602 3122
1529 1732 3300
1529 1666 3431
1529 1766 2933
1577 1348 2901
1590 1767 3401
1590 1549 2981
1617 1740 2995

As you can see it was on the 12th pairing I finally got a match with someone with fewer cups. Also occurred on 14th pairing. That’s 2 of 15 or 13%. A small sample certainly, but these results are alarming.

You can also see a number of cup droppers - 3 or possibly 4 proposed opponents have TS that don’t correlate to their cup total.

I’ll post additional data during my next flask.

1 Like

My anecdotal sense now that I’ve been logging data is that if I’m winning matches, I’ll usually get paired with someone with more trophies thereafter, and after losing a match I sometimes might get paired with someone with lower trophies. There are exceptions to that, but it seems to be the trend.

Since most offensive raids tend to be won, this makes it unsurprising to me that most of my pairings are with players with more trophies, at least until I lose a match.

Next set of data

1617 1783 2995
1617 1627 3336
1617 1715 3463
1617 1490 2775
1637 1636 3214
1637 1595 3182
1637 1591 3209
1637 1861 3281
1637 1849 3350
1637 1776 3420
1637 1706 3410
1637 1823 3246
1637 1868 3365
1637 1913 3346
1637 1367 2816
1648 1672 3439
1764 1975 3382
1764 1558 3230
1764 1883 3419
1764 1926 3437
1764 1503 3057
1775 1603 3111
1792 1868 3477
1792 2009 3534
1792 1753 3288
1792 1800 3400
1792 1946 3518
1792 1831 3421
1792 1687 3249
1792 1837 3230
1792 1627 3226
1792 1977 3436
1792 2020 3605
1792 1862 3405
1792 1882 3268
1792 1952 3386
1792 1895 3209
1792 2032 3422
1792 1880 3205
1769 1927 3588
1719 1800 3363
1719 1833 3448
1719 1725 3344
1719 1737 3321
1719 1527 3077
1734 1982 3366
1734 1936 3495
1734 1799 3130

My cups fluctuated a bit - I tried some revenge attacks which failed. Those are not included above.

A number of teams with lower TS that looked attackable were made up of 2 to 5 5* and I passed on those.

From this data set I had 12 proposed foes with fewer cups out of 48 proposals. That’s 25% which is better than my last data set but no where near the 50% I’d expect if SG isn’t tilting the board.

1 Like