Small Earth has two bonuses with identical value/territory ratings. Except one of them is far preferable to the other. Does the AI always pick that one? Or can it go for the less preferable alternate?

As I said before, the easiest solution is to build in a delay between computation completion and commitment being enabled. Making it 30-60 seconds would practically ensure that autopilot is slower than manual selection, especially since computation time is still factored in.

Solution 1: Make the Autopilot slower.
Solution 2: Remove Small Earth from the coin games, replace it with a sensible template.
Solution 3: Connect Australia to Africa.

If one or more of these does not happen, then the problem will not go away.

Solution 1: Lag everyone's life because of one type of autopilot user. Nope.

Solution 2: It's definetely better than "Tic Tac Toe Reverse" (lol). Small Earth is a great template.

Solution 3: ...

2 and 3 only solve the problem as it presented itself in these specific circumstances but does nothing for potential abuses on other templates

Oy bae, wut you saying about Tic Tac Toe Reverse? My #1 template!

Small Earth has two bonuses with identical value/territory ratings. Except one of them is far preferable to the other. Does the AI always pick that one? Or can it go for the less preferable alternate?

The AI is smart: When only one of Australia and South America is a FTB, the AI always picked the FTB. When one of them is wastelanded, the AI always 1st-picked the other one.

But not always: (1) When both or neither of Australia and South America is a FTB, it's unclear to me how the AI decided which one to pick first. So it might choose the wrong one. (2) When one of AU and SA was wastelanded, the AI sometimes stupidly put a 4th pick in the wastelanded bonus:( (3) Also the AI ignores the euro combo. This can cost a game.

Yes you have the option of undo AI's choices in case it goes wrong, if you opponent has yet finish picks.

Edited 7/6/2016 21:44:22

@ Green Turtle. I was the one who reported you to Fizzer on one of my alt accounts because I thought you were cheating. I am glad we can see the truth concerning this issue (meaning, we now know it was an advantage given by the auto-pilot). I can see now that your behavior wasn't illegal but was still unfair to other players as you were taking advantage of what I would call a flaw of the WL autopilot. It was clearly an advantage for players of WL that are members and that knew how to use the auto-pilot to their advantage.

It is nice to see you are taking actions to stop this unfair practice and that you are trying to find solutions to it.

Sincerely,

Ben

Edited 7/6/2016 22:05:53

@ Benoit. Thanks for your understanding.
Sincerely,
Green Turtle

@Green Turtle, Good Choice in making this thread. I am still curious through as to how you got 0.055 Seconds. Even if you are next door neighbor to the WL server this seems very fast.

Also, 5C of ToS

C. Robots not allowed for game playing.
Users agree not to use or launch any automated system except via warlight APIs defined at http://warlight.net/api that accesses WarLight's website, including but not limited to systems that attempt to automatically play the game, create games, communicate with other Users, or mine data.

Would make Autopilot cheating in my view. Of course, Autopilot is part of WL, and probably this don't cover Autopilot in WL's official interpretation of Robots. After all Autopilot is part of the website. Nevertheless, as have already been said many times, Autopilot and coin games has a bad taste.

@LeQuébécois_Benoit's report was ignored. However, I was unsure about that moderators decision and personally contacted Fizzer. Basically I asked Fizzer in a neutral way, how it was possible to pick so fast. (Before doing this I had tried what OP's instructions tell you to do. This gave me a speed of more then 1 second. (even on truel) Yes, my wifi is not the fastest, but still...) Thus picking manually was faster for me then Autopilot.

So Fizzer responded that "Half a second seems reasonable to me." He also pointed out that it is important to disable https://www.warlight.net/ChangeAnimations For me this brings me down to under 1 sec. Best I have gotten is .7 something, using Autopilot. (I have a better record on 1 spot pick, without autopilot. But never under .5 personally.)

Anyway, Autopilot explains the ~.47 picks consistent picks. Presumably Green Turtle is close to the WL server so lag and latency is small. Regardless, I still have no clue about the 0.055 second game however.

Edited 7/6/2016 22:40:23

@TBest

I don't know what caused that 0.055 either. It is by chance and far in the left tail of my speed distribution. I don't think I can replicate it anytime soon. That's why I was awed by its rarity and decided to frame it into my avatar.

My pick speed distribution using the autopilot is roughly like this (in small earth coin games):
>0.5 seconds by 6%
0.4-0.5 by 90%
0.3-0.4 by 2%
<0.3 by 2% (0.055 once, 0.13? once, 0.174 once, 0.238 once)

Here is the 0.055 game
https://www.warlight.net/MultiPlayer?GameID=11476204

My hypothesis on how I got 0.055.

(1) Hardware: I lived in Bay Area so I guess it's relatively very close to the warlight server. I have a good wifi. My laptop has a intel-core7 processor so AI programs run fast. I used a good mouse and set its double click speed fast.

(2) Settings in warlight: I turned off animation and I checked the box for permitting storage of maps on my hard disk.

(3) Luck: Why is it so hard to reduce my pick speed to <0.4 seconds? According to my observation the major barrier is that the warlight server normally did't register my two clicks within 0.3 seconds. It is often my 3rd or even 4+ th click that triggered "run and commit". My 2nd and 3rd clicks did landed on the overlap area of "begin" and "run and commit", but mostly failed to launch "run and commit".

(4) Malfunction of the speed clock in warlight:
Even if I lucky enough to trigger autopilot with my 2nd click, 0.055 seems very implausible considered that a double click usually takes longer than 0.1 second. Plus the run time of AI, web latency, 0.055 must be beyond the physical ability of my fingers. Thus I am in favor of the "clock glitch" hypothesis below.

In very rare occasions the first turn speed was recorded incorrectly. I encountered this problem at least 3 times (not counting the 0.055). Twice I found a blank entry in the place of my first turn speed. No number showed up. I once boasted to another player that I picked so fast that the system thought I picked in no time:) Another time I noticed a change in my opponent's first turn speed between two visits to the "player panel". He complained that his speed was faster but got adjusted to a slower speed later and consequently lost his first pick.

I cannot confirm whether the 0.055 is caused by speed clock malfunction, but I think the chance is high.

My conclusion is that the 0.055 was caused by a combination of the above factors.

Edited 7/6/2016 23:46:15

The clock can be strange. I had one where times were shown as being identical. After asking Fizzer he said it went down to milliseconds. Not sure how much I trust that...

http://warlight.net/multiplayer?gameid=8868786

Interesting, "Twice I found a blank entry in the place of my first turn speed. "

So... I have actually seen this before, through not from a human player, but from the AI. See: https://www.warlight.net/MultiPlayer?GameID=11521096 (yes, that is me trying to pick fast. ) Here the AI is apparently the slowest player (Cycle position 3.)

Anyway, I presume that is what you mean by empty box?

I also have AI empty box here, https://www.warlight.net/MultiPlayer?GameID=11520147, However in that game <-- I can hoover over the empty box, and it gives me start/end time. (unlike in the first linked game.) This shows that an Empty box, at least in this case, equals a recorded pick time of 0.000.

Can you see start/end time by hoovering over the box? This may actually prove that you "picked so fast that the system thought I picked in no time" :)

Do you have the game link to it?

Edited 7/7/2016 00:07:43

Exactly what I meant by empty box. One of my case happened when I picked manually (my 1-manual-pick has a normal range of 0.3-0.6 seconds)

No you can't measure time properly by hovering over the box. The star/end time are rounded to 1 second. It's common to "picked in no time" in that measurement, when my actual exact time was less than 0.5 seconds.

It's very hard for me to find those links. They are buried too deep in my past games, and I am not sure if they happened before I migrated into this account.