CT Only: Eurisko and TCS

Hi all,

Purely for my own entertainment, I am attempting to learn as much as possible (after 35 years!) about the Trillion Credit Squadron tournaments at Origins 1981 and 1982 that were won by fleets generated by Doug Lenat and his Eurisko computer program.

I have the 1981 fleet listing from JTAS #10, Lenat's original academic paper, and the articles by Johnson (1984) and Gladwell (2009) plus whatever snippets I could find here on COTI and the Steve Jackson Games Forums.

I also have (of course...) 2nd Edition High Guard (1980) and Trillion Credit Squadron itself.

Even so, there is still a lot of information that I wish I could find, but is either not online or beyond my meagre Google skills to unearth. I would be extremely grateful for any further information or hints as to where I might find answers to the following specific questions:

1) Which version of High Guard was used for Origins 1981? The timing would suggest 2nd edition, but nice to know for sure. I have seen it written that the changes from 1st to 2nd edition were in response to the 1981 tournament, but obviously the timing seems to contradict this.

2) Following on from this question - I do not have a copy of the 1st edition High Guard - is it available anywhere, or at least a summary of the changes?

3) Were there any specific tournament rules in effect, beyond those specified in Trillion Credit Squadron? Accounts are divided on this, although without further information my best guess is "no" for 1981 and "probably" for 1982...

4) The 1981 fleet is specified in JTAS #10. Is the composition of the 1982 fleet available anywhere?

Help with some or all of these questions will be greatly appreciated!

Many thanks,

Pete

The FFE CT CD-ROM has both versions, so you can see the differences.

HG'80 was used at the tournament.

There had to be tournament rules or referee made up stuff because it is on record that his fleet would scuttle ships or self destruct in order to avoid penalties - there are no such rules.

kilemall - thanks for the information. My copy of the CD-ROM does not include the 1st Edition of High Guard (or for that matter the 1st Edition of Book 2) but then again it is about 10 years old... I will ask politely if I can get an update

mike - thanks for the confirmation. I agree the "scuttling" sounds like a referee ruling on the day. Lenat's paper implies there were also fleet construction rules that were released very close to the day of the tournament, but it is hard to be 100% sure (especially without more data on the fleet composition). My guess is that some sort of "fleet-based agility" instead of per-ship agility was in force, the question is whether this is simply the Book 5 initiative determination rule or something more restrictive...

Anyhow, a few more useful items, it all helps so thanks very much!

petedavies said:

kilemall - thanks for the information. My copy of the CD-ROM does not include the 1st Edition of High Guard (or for that matter the 1st Edition of Book 2) but then again it is about 10 years old... I will ask politely if I can get an update

Click to expand...

I seem to remember an announcement at one time that the replacement disk was $5 to cover shipping, or free with any order (since the order covered the shipping fee). I know I got my new disk when I ordered some other disks. It included some much higher resolution scans of some books along with the older editions of the books.

kilemall - thanks for your note! It fired a long-dormant neuron, I did some box-searching, and sure enough I got an updated CT CDROM when I purchased T5...

Even a quick skim makes it fairly obvious the fleet specified in JTAS #10 used 2nd edition High Guard (although I will check in more detail later).

Ok, reading through this I remember a number of conversations covering this topic involving Loren. Unfortunately He would have been the person to ask.

But if I were going to go look, remembering/having heard of conversations from 30 years ago I would go poke around the early archives of the TML and/or GEnie and Maybe the archives of CT-Starships on Yahoo.

Sorry, I don't have URL's for any of those sources.

Evyn - thanks for the pointers, I will try a bit more searching along those lines.

I did find an interesting comment from Loren on the Steve Jackson forum, every little helps

This paper by Lenat might be of use. It discusses Eurisko's inner workings and gives a fairly detailed look at how it was used for TCS.
http://www.cs.northwestern.edu/~mek802/papers/not-mine/Lenat_EURISKO.pdf

My problem with the Eurisko Incident is that the initial fleet of
"super-rocks" should never have been ruled qualified to enter the TCS
Tournament due to egregious drop tank abuse and the resulting failure
to realistically meet the Jump requirement... but that's all water
under the bridge (or maybe L-Hyd under the gantry)... as mentioned,
the designs were superior not so much from all the massive redesign
iterations to find the optimal MaxMins in the HG2 rules (which trial
& error have borne out for most of us anyway over the years), but
rather from the fact that they violated the principle, if not the
letter, of the contest performance parameters.

Truly, Eurisko is the god of Munchkinizing and Rules-Lawyering... not AI...

Eurisko is the god of Munchkinizing and Rules-Lawyering... not AI...

Click to expand...

what's the difference?

It's legal to pull the plug on an AI - you get into bother if you 'pull the plug' on a munchkin or rules-lawyer

Ishmael - thanks for the link. Unfortunately Lenat's paper does not answer the specific questions I'm interested in (that's why I asked them here, really)

jec10 - not sure I share your point of view regarding drop tanks, although of course in a referee-controlled game theirs is the only one that really counts! In a tournament, if you design to the rules as written (and reading HG2 and TCS, it seems Lenat's fleet does that) then it seems fair to me. After that, his fleet (IMO) just applied Lanchester's Law (or even the "Salvo Model", albeit 14 years before that was created!) to do the job.

Of course, the "lifeboat" concept used to win the 1981 final round is *extremely* gamey, and bears no relation to any sort of "reality" - but still, if you read the rules in force on the day I think you can make a lawyer's case it's allowable. My 2 cents is in an ideal world that would have been disallowed by the ref and the 2 final fleets allowed to go head-to-head, in which case Lenat may still have won - but that's all history now.

Experience would allow tweaking of points and victory conditions.

The surviving vessels would need to secure certain objectives, which would require specific performance parameters.

Condottiere - exactly! That's why I'm so interested in the 1982 tournament. The 1981 gets all the detailed references, but what intrigues me is how the rules were adjusted for 1982 and how the Lenat/Eurisko fleet was completely redesigned (and won again). To be honest I don't have a strong opinion either way on the fairness/"realism" etc aspect, it's just an interesting case study in game design.

It would be useful to hear about rules changes (if there really were any, opinions are divided on that) but I suspect you could work out a lot from the composition of the 1982 fleet - I'm really curious about that particular question!!

infojunky said:

Ok, reading through this I remember a number of conversations covering this topic involving Loren. Unfortunately He would have been the person to ask.

But if I were going to go look, remembering/having heard of conversations from 30 years ago I would go poke around the early archives of the TML and/or GEnie and Maybe the archives of CT-Starships on Yahoo.

Sorry, I don't have URL's for any of those sources.

Click to expand...

I might have some of those discussions.

Try my site, under:
==> The Best of the TML
==> Starships
==> Starships (HG) - Professor Lenat and EURISKO's Winning Fleet

Or the direct link:
http://members.tip.net.au/~davidjw/...ofessor Lenat and EURISKO's Winning Fleet.htm

Hi David,

Thanks very much for the link. As it happens it is discovered by Google so I had already seen it (in fact it was reading that exact thread that re-ignited my interest in the whole subject!) but still, much appreciated.

To continue the analysis that was happening therein - my understanding is that the 75 "Eurisko" class were the main battle-line, optimised to exploit the hitting/damage rules in HG2 (analogous to "human wave" tactics).

The "Wasp" class is what Lenat referred to as a "lifeboat", which enabled the (extremely dubious) 1981 tactic invented to defeat any opponent who used the same core design philosophy.

The "Queller" class is his "lifeboat killer" design, as a backup plan should an opponent also have "lifeboats".

The other two classes I don't understand yet - but I think it likely one of them exploits some other pathological corner of the rules and the other is the countermeasure. It will be a fun exercise to try to deduce what that could be

After all this time, this may be the best I can do - but I will keep on searching for that elusive 1982 winning fleet! I do have a tentative theory of how it operated - but that is *highly* speculative, since based on only a few tiny scraps of data in the published sources.

Pete

jec10 said:

My problem with the Eurisko Incident is that the initial fleet of
"super-rocks" should never have been ruled qualified to enter the TCS
Tournament due to egregious drop tank abuse and the resulting failure
to realistically meet the Jump requirement... but that's all water
under the bridge (or maybe L-Hyd under the gantry)... as mentioned,
the designs were superior not so much from all the massive redesign
iterations to find the optimal MaxMins in the HG2 rules (which trial
& error have borne out for most of us anyway over the years), but
rather from the fact that they violated the principle, if not the
letter, of the contest performance parameters.

Truly, Eurisko is the god of Munchkinizing and Rules-Lawyering... not AI...

Click to expand...

TCS: "The parameter statement indicates the minimum jump drive performance required for the squadron. The parameter states that the squadron must be capable of a certain jump level; not every ship in the squadron must be fitted with drives to meet or exceed the required level, but fleet tenders or fighter carriers must be included which are capable of carrying those ships and craft which not fitted with the required jump drives."

"1981 Tournaments: The maximum tech level allowed for all ships is 12. The
total pilot allowance for the squadron is 200. The squadron must be capable of jump-3; each ship and small craft must be capable of 1-G acceleration. The squadron must be capable of gas giant refueling."

Reading the guy's article, it looks like the idea to use drop tanks was his rather than the program's. It did in fact comply with the rules as written. His squadron was capable of jump-3 and carried the fuel needed to reach the battle site. Nothing there says he needed to be able to leave once he got there - and arguably he could have had freighters jump to the outer system to bring him filled replacement tanks if he'd tried this in campaign play. There is no "realistically" or "unrealistically" in convention tournaments, not for an abstract battle game with an infinite supply of missiles: you either write your rules and conditions tight or some rules lawyer will slide in through a loophole. Closing a loophole on the argument of "principle" or not being "realistic" in a game with infinite missiles tends to draw accusations of bias and gives organizers a bad reputation unless such a requirement is declared in advance, since that argument could be used against any number of clever ideas.

According to the guy's article, the number two fleet used the same heavy armor idea. Probably the same smother-them-in-missiles tactic. Didn't say how that guy handled fuel. What'cha wanna bet he used tenders? Opponent yielded because he didn't have an unhittable ship to hide behind for repairs, nor an unhittable-ship-killer. The rest of the players used conventional spinal mount battlewagons, according to the article.

He hinted that his 1982 fleet flipped the script and went all offense instead of all defense. I have to admit that tickles my curiosity too.

Carlobrand - I 100% agree with your view. I think the core of Lenat's 1981 fleet is aligned with the letter and spirit of the HG2/TCS rules, and proved once again the effectiveness of Lanchester's law under such conditions.

The "lifeboat" tactic, IMO, is a loophole exploit that is way over the line

I haven't spent any time checking the impact of 1982 conditions on fleet structure (TL13, more pilots, etc) so it's hard to say if missile swarm is still optimum. I am almost certain it will still use the "swarm tactics" idea though.

The really interesting comment (to me) is around the importance of fleet agility. My speculation is as follows:

1) The tournament refs outlawed the "lifeboat" tactic, probably by tweaking or eliminating the "repair" rules during an actual engagement.
2) Based on this, and the fact that the 1981 tactics were known, the 1982 fleet would have to contend with opponents that used the same design philosophy, and couldn't hide behind a "lifeboat". Therefore the performance would need to be optimised against such a fleet.
3) Whatever offense/defence loadout was chosen, it is highly likely that optimising range each round (long/short) would maximise performance. Fleet agility is a major factor in being able to dictate this...
4) I speculate - the 1982 fleet was optimised for combat at short range (massive amounts of lasers/energy weapons), the agility and ship count were maxed in order to maximise the chance of choosing optimum range each round. The "new exploit" was to self-destroy any damaged ships with reduced agility, to keep this advantage as the battle progressed.

I am probably wrong - but I would love to see the 1982 fleet to see for sure!

P.S. The 1982 fleet was also probably lightly or negligibly armoured, based on the "offensive" versus "defensive comments --- maybe