Deconstructing the Galactic War Power Curve: A Detailed Look

I originally posted a discussion about the GW power curve on 2016/06/30. Thread is here and recapped in the spoiler below.
https://forums.galaxy-of-heroes.starwars.ea.com/discussion/50621/sorry-everyone-i-broke-gw-todays-power-breakdown
I ran an experiment to determine how GW nodes were scaled: was the opponents power scaled based on the player's best Arena power, the player's 5 most powerful heroes regardless of Arena (I'll call them the Top5), or something else? EA/CG changed the algorithm on me twice during the experiment, so that limited the amount of data I could collect.

During the experiment, I held my Arena power and Top5 constant. Some of my Top5 were in my Arena team, and the top of my roster was fairly equal so my Top5 and Arena power were almost the same at ~35k. Not an ideal situation, but that's how it was. After GW was retuned on 06/30, I determined, by node, the new power curve based on Arena power was:
  1. 60.0%
  2. 62.5%
  3. 70.1%
  4. 61.1%
  5. 64.5%
  6. 95.5%
  7. 76.2%
  8. 77.0%
  9. 100.4%
  10. 86.3%
  11. 104.1%
  12. 115.4%

That curve was calculated on 06/30 and confirmed on 07/01.

As it turns out, the curves in the previous post were a little off because I was eventually able to determine that Top5 is more important to GW scaling. So, let's go over it all again, shall we?

Warning: there's a lot of data below. Peruse at your own risk.

A. GW Power Curve, pre-06/15
Based on Top5 power levels:
  1. 60%
  2. 65%
  3. 70%
  4. 77%
  5. 80%
  6. 84%
  7. 90%
  8. 95%
  9. 96%
  10. 97%
  11. 97%
  12. 100%
This part of the study didn't actually prove anything sadly, but you might be interested in seeing what we used to deal with.
  • Arena: 34883
  • Top5: 35047
I started collecting data on 06/08, although I wasn't doing a great job at first. During this period I managed to collect data from 8 GW runs. I completed all of them. Here's the power levels for those runs, both mine and my opponents:
KEPYuY9.png
You can see I was a bit sketchy at first, and clearly did an upgrade to my Arena squad on 06/10, but the power level was constant from then on. I don't recall what happened on 06/14 to boost the Top5 by 30 points, but this was a small change and I'm just going to ignore it.
If you convert those numbers into percentages based on Arena power levels, you get the following curve:
HRTZHCt.png
Converting into percentages based on Top5 power, you get:
SjJEZ1h.png
The curves are basically the same and we haven't proven anything yet. Then CG retuned the GW matching algorithm on 06/15 and screwed up my experiment this work went out the window.

B. GW Power Curve, 06/15 - 06/29
Based on Top5 power levels:
  1. 60%
  2. 62%
  3. 70%
  4. 61%
  5. 62%
  6. 80%
  7. 71%
  8. 72%
  9. 84%
  10. 81%
  11. 99%
  12. 100%
This part of the study didn't actually prove anything either because CG retuned GW again on 06/30. I was still collecting data to baseline the curve. But here it is
  • Arena: 34883
  • Top5: 35047
During this period I collected data from 14 GW runs. I completed all of them. Here's the power levels for those runs, both mine and my opponents:
yCAiaI1.png
Still some holes in the data because sometimes I just forgot to record it. One funny looking number shown in yellow because it just doesn't fit. Maybe I added wrong, I don't know and I honestly don't care anymore.
Percentages based on Arena power levels:
bMnsOUp.png
Percentages based on Top5 power:
pMlyL0W.png
Again, the curves are basically the same and we still haven't proven anything yet. After the 06/29 run, I decided it was time to power up one of my non-Arena Top5 and find out what was what. So I upgraded Savage Opress from L77-IX to L80-X, increasing his power from 6779 to 7540. This of course, coincided with CG's GW retune on 06/30, the one they didn't tell us about.

C. GW Power Curve, post-06/30
Based on Top5 power levels:
  1. 59%
  2. 61%
  3. 69%
  4. 60%
  5. 61%
  6. 93%
  7. 74%
  8. 75%
  9. 98%
  10. 84%
  11. 101%
  12. 112%
This is it, the one you've been waiting for. Proof! Kind of.
Having learned my lesson from the last two tries at this experiment, I started ramping up my Top5 after the first 9 runs to see what would happen. Guess what? Things started to change.
Now, it's critical you understand this: Arena power was constant throughout this period. If GW were based on the max power level of my Arena wins, then increasing my Top5 power should not have affected my GW opponents. It did, therefore, GW is scaled based on Top5, not Arena.
  • Arena: 34883
  • Top5 (ending): 36468
During this period I collected data from 13 GW runs. I may have actually completed two of the runs; I know I finished the run on 07/11 when I got an easy node 12 (that's the only one I've ever seen). Here's the power levels for those runs, both mine and my opponents:
c0c6F1j.png
A couple of data holes but not bad. 07/12 was Mod update day. I couldn't get past 9 on that run. After that, I was just disgusted with the whole GW thing and quit the experiment.
Percentages based on Arena power levels:
2Pr9bep.png
Percentages based on Top5 power:
yhA6KtY.png

Hold on a second. Before you math freaks point it out, the averages and standard deviations shown here are only for the GW runs through 07/09. Things really started to go haywire as I upgraded my Top5 after that point. You can see it in the percentages: check out Nodes 2, 3, 5, 6, 7, 9 ...
That craziness in the percentages just added a more questions.
  1. Are the node % levels really what I've shown above? They're varying so much, it's hard to know.
  2. Perhaps the percentages aren't based on Top5, but instead some other metric (Top6? Top10?). I could explore this concept more thoroughly if I wanted to, I have all the records, but I just don't care enough anymore.
  3. Maybe the matchmaking isn't based on straight percentages. If it isn't, then what is it based on?
  4. If matchmaking isn't based on percentages, they why do I reliably get the same percentages when I hold power constant? If it's based on some other algorithm, one would think the standard deviations would be larger.
  5. Was CG retuning GW difficulty yet again during my experiment?

So there you have it. The power curve shown in section C is probably close to correct, or at least was at some point. I may not have been able to nail it down exactly, but one thing I can say 100% for sure:
GW node scaling is definitely NOT based on Arena power, but is instead more affected by the most powerful heroes in your roster.

Now that I regularly can't get past node 9 anymore (sometimes 6), I am too fed up and disgusted with the GW to put anymore effort into it. For the record, I am currently sitting at
  • Arena: 36749
  • Top5: 37702

And that's it from here.

Cheers

Replies

  • I am close to your numbers and running into the same problems. Used to be my favorite part of the game, now I haven't finished since the update.
  • I finish in most cases. Except when I get the team without a healer that regenerates health or the team that attacks 12 times before I get 2 then they get another 12...
Sign In or Register to comment.