This page is for discussing scoring and rule changes for future Crawl tournaments.
There was a lot of support for making the tournament shorter. (Also, the usage statistics for August showed a significant slump in playing after two weeks, for all tournaments so far.)
The current (2010) tournament has the rule:
100 extra points for each consecutive win by a player that is not a repeat race or class of any of the wins in that particular streak of wins.
A problem with this rule is that it encourages players who can win reliably with a few strong combos such as SpEn and DDNe, but not as reliably with mediocre combos, to play these same strong combos repeatedly at the beginning of each streak. Speaking as such a player, I would prefer to play a wider combination of combos, but that is disincentivized by the scoring. Another problem is that a player who can win 2 or 3 combos quickly and reliably can play a series of streaks, gaining 100 or 200 points from each, limited only by availability of playing time.
There are probably several ways to remedy this. Here is one suggestion (by — dpeg 2010-09-07 15:01)
This should shift emphasis from streaks to unwon species/backgrounds which is probably more interesting.
I have pruned two older proposals, to which commentators pointed out problems:
Streaks are the greatest show of skill in my opinion. However, you should not be rewarded on multiple streaks for the same characters. If you've done a SpEn in one streak another SpEn in a second streak doesn't show that much skill and shouldn't be worth the 100 points. Unwon characters makes sense except that the tournament is short (ie not several years long, for good reason) and so there will be many unwon combos. If unwon (and not streak) combos are the norm, this will happen; say OgWz is the easiest combo. Someone will play lots of OgSt (unwon) until they find an early book of conjurations and then win. This shows less skill than elliptic's 10 streak of different race/chars which shows the ability to win whatever luck you get. Streaks should be the most important things (as now) however you should not be able to win multiple SpEn/DDNe/HaBe to rack up the points. — casmith789 2010-09-10 17:12
The way streaks are determined currently could be improved. If you only play on say cao, you're on a streak and your game goes bad (say 10hp in Zot or poisoned by a snake on D:2), you should save your game (don't end it!) and start a new character on cdo. You can then continue your streak until you die in a cdo game. — rob 2010-09-03 17:23
The 2010 scoring of 50 points for first time finding each rune seems rather high. For a player who already has two wins (to eliminate effects from first/second win points) winning an all-rune game is worth about the same amount as five 3-rune games won on streaks. Also, a player could reasonably find six runes over the course of several games without ever winning, which could be worth more than a single 3-rune win. I like the increase from the 2009 value of 10, though; finding runes is a significant milestone for intermediate players.
Another issue is the binary nature of the all-runer bonus: Right now there's very little reason to do anything but a single all-runer then nothing but 3-rune wins for the rest of the tournament. elliptic suggested in IRC: “the simplest way of fixing the lack of incentive for allruning in tourney would be to make every allruner worth an extra 50 points, not just the first one”. I think something along these lines would be a definite improvement to the rules for next year. — marvinpa 2010-08-24 00:35
dpeg 2010-08-16 00:39: Instead of awarding 'pure god' wins, I suggest to
Hand out the 20 extra points if you manage to reach ****** piety with a new god.
The advantages are that
(This was already planned for 2010, but didn't make it in for time reasons. I only added this here lest we don't forget.)
+1. However, for Xom, the system should stay the same, and perhaps force the player to pick up xom in the first 10000 turns (so you can't pick up xom on the way up of a godless run). — casmith789 2010-09-10 17:07
The choices are very good, but they should be stressed some more:
I was thinking the next nemelex choice is unveiled as soon as the first player wins a nemelex choice. Each nemelex choice is worth 50 more than the last and all must be won in order (but not on streak). The first nemelex choice is worth 50? (maybe 100) points, so as soon as a GhFE is won 50 points are got from it, then 100 points from a SEAs after that etc. (obviously different choices for next year) This would allow many nemelex choices but in reality only a few; especially with a shorter tournament. — casmith789 2010-09-10 17:04
The number of kills counted for the purpose of determining most pacific win include toadstools killed by either the player or monsters (who might just be wandering into them). In one of my games there were 1868 monsters killed, of which 396 were toadstools killed by hostile monsters. One can avoid toadstool generation by worshipping a god who accepts sacrifices of corpses or by chopping up every corpse. Either way, this encourages tedious (and potentially nonobvious) play. It would be better to not count toadstool deaths in the kill count, and maybe to extend this to other plants.
dpeg suggests to change clan scoring to the following model:
Treat a clan like a single player. Don't care about streaks but take care of gods, runes, high scores (only compare clans with clans).
We don't want banners which induce tedious gameplay or long preparation. I don't think we want banners that are purely random (e.g. best artefact). Here are two ideas dpeg likes:
On this note, I would recommend against banners (and point rewards) that encourage players not to win. For example, the Jiyva banner asks players to destroy the Orb of Zot, and the first time they do so is 100 points. Since this is not a win condition in Crawl, after winning two games and not being in a streak, it is worth more to not win! zannick 2010-08-26 19:28Banners like the Chei (hare) banner should not be banners. It is a good mechanic but only one person can get it; Beogh maybe also. All banners should be in reach for all people imo. I also disagree in principle with the royal jelly banner; it is effectively rewarding someone for playing (and effectively winning) the easiest combo available for a second time with no real skill except the orb banner. — casmith789 2010-09-10 17:08
We unleashed a new Sprint after two weeks in order to keep players motivated. But perhaps that's not such a good idea, and it'd be better to shorten the tournament (duration) and to stage separate Sprint tournaments. Ideally, interested players would organise them. The scoring code is available at http://github.com/greensnark/dcss_tourney. Some vague ideas of what could be done:
I have finished an initial draft of the new rules for the August tourney (including a full list of banners), available for viewing at http://crawl.seleniac.org/t11/ for anyone who is interested. I will probably not have time for anything Crawl-related for most of the second half of July, but I will certainly take into consideration any comments left here when I finish up with updating the tourney scripts at the beginning of August.
Also, the new banners (9 of them) need images! So I'm now soliciting contributions for that from anyone who is interested in helping that way. The banner names and descriptions are on the rules page linked above, along with the old banner images (those that are still being used) to compare against. — elliptic 2011-07-12 10:55
New to the wiki, but have been participating in the tournaments since '09 and always look forward to it. Thanks for putting this together, I and many others appreciate the work you all put into this! A few comments:
Thanks for the comments! I hope I can clarify how a few of the new rules work (and the reasons for introducing them).
Great clarifications, you've allayed most of my concerns. I have one question about Nemelex's Choice below.
* I'm not adamantly opposed to adding a speed component to everyone's score, but points for turncount and real-time seems a bit much. Honestly, I would prefer all real-time duration points (top 3 included) to be included only in clan scoring, since it's not part of the traditional roguelike skill-set, in my opinion; also, end-game score does not take real-time used into account, which is another argument for this being a specialty skill like other clan-scoring components. The proposed setup seems to put even more emphasis on who is best with SpEn, and there's already a fair amount of emphasis on that in the previous tournament scoring rulesets.
Giving everyone points for speedy wins was actually intended to decrease the significance of speedruns in the overall ranking, since everyone will be getting significant points from it (even people who aren't trying at all – a 100k turn win is worth 50 points, etc) rather than just the few people who are best at speedruns racking up many hundreds of points and everyone else getting nothing. Also keep in mind that speedruns have been made substantially more dangerous in 0.9 because of the changes to the orb run. The constants 5M and 1.25M might well get tweaked a little, but the intention is definitely not to emphasize “who is best with SpEn”… this is just one of many ways of getting points. Also, note that each player only gets turncount points from their fastest game… previously players could rack up a lot of points with multiple speedruns dominating the top 3 list(s).
The other reason for this change was that a commonly expressed sentiment in the last couple of tournaments was “There is no point in my trying to speedrun, since players X/Y/Z will just do faster ones” – and I think that it is better if components of the tournament are accessible to more players than just a few. This certainly doesn't mean that I think everyone should feel forced to try a speedrun (there are plenty of other ways of getting points, and if you aren't good at speedrunning then the others will definitely be more efficient), just that people who feel like doing so (and succeed) should be rewarded a little bit even if they aren't the fastest players.
I didn't realize that this replaces the top 3 speed awards! OK, that's much better, I thought it was just in addition to the top 3 speed awards. I haven't experienced the 0.9 orb run yet, as I prefer to play stable most of the time, but I'll be interested to see how that affects Spriggan speedruns. Feeling a lot better about this, thanks for the clarification!
About turncount vs realtime… people seem pretty evenly split on the question of which one is a better measure of speediness, and they are certainly both tests of skill, so I'm reluctant to treat them that differently from each other.
* I'm not sure that Nemelex's Choice fits in with the component of the scoring that penalizes popular winning species and backgrounds. Won't the bonus points for Nemelex combos be negated by receiving less points for using a popular species/bg (since at least a few players will win with it)? Unfortunately, I'm not sure what a good solution incorporating both these components would look like… maybe exclude Nemelex combos from the species/bg popularity score? That makes an already complex calculation even more un-intuitive, though.
Yes, this anti-synergy was the main reason why there was no Nemelex' Choice in the May tournament… I couldn't figure out a reasonable way of making it work at the time. But the new rules (both for species/bg points and for Nemelex' Choice) will hopefully work together better for the following reasons:
* Each Nemelex' Choice is closed after five people have won it, so there will only be four extra previous games at most… not a huge number (consider that every species/bg was won at least five times in the May tourney).
* The species/bg points only take into account games won before the given game started… so if you start a Nemelex' Choice combo and then three other people win that combo while you are playing, that doesn't hurt you.
* There will always be a fresh Nemelex' Choice option, not yet won by any players.
One question here. say I start a Nemelex combo when three of those have been won already. two more people win the Nemelex combo during my game. Is my game no longer eligible to get the banner/bonus?
Yes, only five people can get the banner/bonus points for each Nemelex combo… so if 3-4 people have already won it, you might be better off starting a newer combo. — elliptic 2011-07-21 22:11
* I notice that the proposed species/bg popularity score only takes into account past games. Is the intent here to prevent scores from shifting around like in the May tournament? I think I'd prefer the May system, since this will give a better overall picture of the 'easy' species/bg wins due to larger sample size, and I think that advantage outweighs the uncertainty of slight score shifts. The proposed system would also exacerbate the penalty for Nemelex games as discussed above, since a lot of people will be playing the same combination at once. — oddsox 2011-07-18 23:17
Yes, there were many complaints in May about scores shifting around a lot. As an example of the sort of thing that happened, minmay was the only player to have won a DE for most of the tourney, so he had about 200 points from that at one point. But then several other people won DEs near the end, taking about 150 points from him. Even in less extreme situations, people still didn't like how difficult it was to determine how much a given win would be worth at the end of the tournament.
Again, about the Nemelex games, if people are actually playing the same combination at the same time then there won't be a penalty for that with the new system, since it only looks at games finished before the given game started. So in the situation where DEDK is revealed as the next Nemelex' Choice and several people start one then, those games would use the pre-Choice point values for DE and DK. — elliptic 2011-07-20 22:54
Makes sense, will be interested to see how this affects scoring in tournament play. — oddsox 2011-07-21 14:53