AI vs AI: Chasing team always wins?

[Jay OTT]

I recently started a 20/20 draft league for DBC on another forum, and were at the part where I play out the games, in this case via watching the CPU duke it out with itself and keeping track of the league table manually. It's been very well received so far, but some of us have a problem with the results that are getting played out...

The chasing team has won all four games so far. And not just won them in close contests, but very comfortably too. The closest any team batting first as gotten so far was losing with five balls remaining, and even that run chase never really looked to be in doubt. The other three margins of victory have been 19 balls, 12 balls, and now 35 balls after the team batting first got bowled out for 121, on a pitch that started the game very green but flattened out all too quickly...

The AI is on Legend difficulty, weather has been set to Random - No Chance of Rain, and pitch conditions have also been set to Random, with normal pitch wear. Is the fact that the chasing team is winning easily just a bad co-incidence, or does the AI in this game legitimately play better when chasing a target than when they're setting one?

 

[Biggs]

Not something I've seen (or to be fair, tried) but whenever I've simulated in career mode it's always been fairly mixed. Might I suggest a larger sample size than four games before making a conclusion?

Your first mistake was posting in another forum, in all honesty...

 

[TerryAlderman]

I never even knew that you could have the ai play a tournament. One thing I have noticed that there is no point trying to get an ai team out for few runs. Like I've been bowling and had teams on for example 100 for 8.....and the last two ai batsmen seem to get the score to a fairly respectable 150-170!

 

[Biggs]

...well when you think about that, that's fairly realistic. It's not often teams are rolled completely in cricket when they're 7 or 8 down. They usually put together a faintly defendable score, especially in T20's

 

[TerryAlderman]

...well when you think about that, that's fairly realistic. It's not often teams are rolled completely in cricket when they're 7 or 8 down. They usually put together a faintly defendable score, especially in T20's

Or scripted?

 

[Markkkkk]

Or scripted?

Shh not so loud! Some people get defensive when you talk about DBC like that...

 

[cooks1st100]

 

[Jay OTT]

You can call this particular 'scripting' issue busted as game five was a wonderful game where the team batting first made a modest 8/130 on a worn out pitch then bowled out their opponents for 125 to win it.

I'll keep an eye on this potential issue as the tournament continues and let you know what I'm observing in AI behaviour in general in a couple of weeks time.

 

[Biggs]

[HASHTAG]#ScriptedRandomness[/HASHTAG]

 

[Langeveldt]

Are you finding that the skill levels of the players correlate to what you see on the screen?

I've watched many AI vs AI matches, have them on in the background while I work, and for me there is absolutely no correlation between the skill levels and what you actually see in a match. Out of ten AI vs. AI matches recently watched, the chasing team has won eight of them. So although not always, there is a definite advantage batting second.

If you want to test it yourself, send in two tailenders to open and they'll score what seems a predefined about of runs before scooping the ball to mid on or mid off. It's like they have to get their personal target, and then they get themselves out.

It's a shame there is no realism in AI vs. AI, as the feature was implemented and it looks really cool, it is also important in career mode of course.

 

[Jay OTT]

Tail-enders are sometimes able to come out and score boundaries from ball one, which I think we can all agree is unrealistic... but they're rarely able to get a big score before running out of talent and putting one down a fielders throat. Results so far are inconclusive, but I do think that the established batsman play better than the bowlers on average. The one time they didn't was in a game where the pitch started out very green in the first five overs, but quickly flattened out after that allowing the lower order to salvage something of a total, only to see that total chased down quickly because the pitch was now a decent batting track.

I've got a total of 42 regular season games to play out, so over the next couple of months I'll be able to get a better sample size and see just how advantageous batting second in an AI vs AI game is, and whether batting skills really do make a significant difference in regards to batting averages...

 

[Biggs]

Tail-enders are sometimes able to come out and score boundaries from ball one, which I think we can all agree is unrealistic...

Tim Southee would disagree, he specifically does a lot of training in the nets to hit boundaries from ball one, Nathan McCullum does the same. I've watched them. They do skill-based stuff around eye co-ordination and hitting boundaries from ball one. The modern game and T20 specifically is challenging a lot of these pre-conceptions about tail-end batting.

Just playing devils advocate, for the sake of it. The AI is far from perfect in the game, but there is a lot of tail-end talent these days that do a lot of work with batting coaches that is essentially "slog training"

 

[grkrama]

SL 67/3 (18.1 ov, AD Mathews 22*, JK Silva 24*, R Ashwin 0/4) - Stumps | Live Scorecard | ESPN Cricinfo

this is a classic example of DBC AI match engine. Lower order kicking it rapidly after top order gets bundled.

Its just ahead of the curve

 

[Langeveldt]

If you want to see the irrelevance of the AI vs. AI match engine, put Johnson and Harris to bowl on a green seamer, and have them up against Russia, and watch how easy it all looks to bat.

 

[Jay OTT]

This issue ended up being a non-issue in the end. In fact with the team batting first now winning the last four (overall score is 6-5 to the chasers), if this stretch of games had happened at the start of the tournament this thread may have been titled "team batting first always wins". Just tonight I saw a game where the chasing team needed just 34 to win off five overs with seven wickets in hand. They proceeded to go 1/9 in the next three overs and turn that into 25 needed in 12 balls, and they went on to lose by two runs with six wickets still in hand.

The AI's tendency to hit it straight to the fielders and conservative approach to running between the wickets tends to result in a lot of dot balls even in the final overs of the innings. So whilst the overall run rate for the tournament has been fairly realistic, there hasn't been any bat-first score of more than 183 so far (and that was on a ground with short boundaries), and post-powerplay overs have frequently bogged down batting teams even when they've made 50+ in the first six overs whilst only losing one or two wickets at most.

And yes, lower order batsmen playing a bit too well continues to be a concern. The fact that they have less skill than the established batsmen is noticeable, but probably not significant enough considering the typical gulf in batting average between established batsmen and lower order bowlers in real life cricket.