Open Side Menu Go to the Top

07-17-2012 , 08:31 AM
If I'm gonna make a play for a win, it'll be in the setup rounds. Like yesterday. I think my Round 4 play was probably the maximum score EV, but compared to the other play has a lower expected head-to-head winrate but a higher overall winrate. So there you go, that's me sacrificing rating for win EV.
Triplechain Quote
Triplechain
$25m Guaranteed WPM on CoinPoker
Join the action now
Daily Rewards • Splash Pots • CoinRaces
Triplechain
07-17-2012 , 10:31 AM
Quote:
Originally Posted by TH10
Yesterday's winning score was 18848, but it was possible to score 45425 in that game, should be close to the worst winning score in comparison to what was possible.
Our boards were scaringly similar in that game

I am on my way to lose daily expert status, chaining one trainwreck to another.

I must say the game is starting to lose my interest.
Triplechain Quote
07-17-2012 , 12:21 PM
Quote:
Originally Posted by Gabethebabe
Our boards were scaringly similar in that game
Indeed.

Quote:
Originally Posted by Gabethebabe
I am on my way to lose daily expert status, chaining one trainwreck to another.
You tried reviewing your setup strat?

Quote:
Originally Posted by Gabethebabe
I must say the game is starting to lose my interest.
+1. I'd like to play more than one game a day, but as it is I just feel like I'm very slowly grinding out towards that #1 in wins and possibly catching TH10's rating, which is really dull.
Triplechain Quote
07-18-2012 , 09:55 AM
SIGH at today's DC.
Triplechain Quote
07-18-2012 , 07:52 PM
if this is unlucky number 13 then I've bloody waited too long for it
Triplechain Quote
07-19-2012 , 08:21 AM
ABOUT ****ING TIME!!!!

first win since the blackout

Booya

see you in 2 months for my next win.
Triplechain Quote
07-19-2012 , 09:16 AM
congrats lord DonkDonkDonkDonk
Triplechain Quote
07-20-2012 , 11:10 AM
Quote:
Originally Posted by DonkDonkDonkDonk

see you in 2 months for my next win.
Wow those 2 months flew by!
Triplechain Quote
07-20-2012 , 11:32 AM
wait an age for one to come along, and two come at once obv.
Triplechain Quote
07-21-2012 , 04:59 PM
In an attempt to try raise interest in puzzles, here is a puzzle solution walkthrough to show one way of solving puzzles. Obviously it will spoil the answer for this particular puzzle (Vehicula). This isn't a very difficult puzzle imo, actually it pretty much plays itself. I should probably do what Gaby did, and give my puzzles some sort of difficulty ratings.

This should be obvious, but just in case: in puzzles there is absolutely no need to play the chains in order, instead you need to plan ahead.

Spoiler:
So here are the racks for the puzzle:


We should start the solving by looking at the amounts of different dice.
These are listed (racks 1-4)+(racks 5-8)+(rack 9):

1s: 4+2+1, no more than 2 in any single rack -> shouldn't be any problems getting full bonus from 1s; nothing to worry about other than leaving one zone free of 1s for 9th rack
2s: 4+3+1, no more than 2 in any single rack -> shouldn't be any problems getting full bonus from 2s; really standard one to each outer zone in first 4 racks, pair 3 in racks 5-8, last one in rack 9
3s: 3+7+0, can be snaked within first 8 racks playing many different setups, can't be snaked with 9th rack, there are 7 in racks 5-8, might cause some problems trying to get them all chained
4s: 6+1+2, can be snaked, but only with 9th rack
5s: 2+3+0, can be 4-zoned within first 8 racks, first two 5s are in different racks -> can be 4-zoned in a lot of ways
6s: 1+4+1, can be 4-zoned within first 8 racks or with 8th rack, however doing either would prevent two other chains from being snaked

So, bonus dice shouldn't cause any problems, let's have a look at the chain dice then. The order of value seems to be 4s>3s>6s>5s, but 6s can't be 4-zoned while snaking both 3s and 4s, so best option would seem to be snaking 3s and 4s, 4-zoning 5s. But then we notice that 4s can't be snaked without help from 9th rack, and at the same time that 6+1+1-snake setup means that we can't snake one other chain while 4-zoning the other. That leaves 5s out of the equation. Then we need to figure out if it is better to 4-zone both 3s and 6s, or snake 3s, and 3-zone 6s, and with this many 3s, snaking them scores a bit more. We already know there is no problem with 2s. With 1s have to remember to leave one zone free for 9th rack, and it can't be the center zone since it is needed for a 4. Also with the 6+1+1-snaking, and snaking 3s too we can't play a 1 to center in first 4 racks (needed for chain dice), but can do so in racks 5-8, 1s in racks 1-4 should be played to outer zones 2-1-1-0 or 2-2-0-0. So in conclusion, we can get full bonus while snaking 4s, snaking 3s, 3-zoning 6s. Then we just have to be careful, and make sure we get all 3s and 6s chained, 4s won't cause any problems with the 6+1+1-snaking. So this was a pretty easy solution in the end.

Usually the solutions don't go quite as smoothly for more difficult puzzles, instead you'll have to make some adjustments. Also, if there are lots of 1s/2s, it can be a good idea to first see how you can play them. Then build chains around 1s/2s, and adjust 1s/2s if necessary. Free move mode is a great tool for this, of course in the beginning it can be helpful even when playing easier puzzles, such as this. Few things to remember with it: 1) Keep the puzzle racks open when using free move mode, sometimes your plan might not be possible to complete with the racks the puzzle has, even though the dice amounts are correct. Use either two different types of browsers or take a screenshot of the racks and have it open on the background. 2) Do not complete the game free move mode, and only place 44 dice at most, otherwise the free move mode score will show up in Recent Scores, and can possibly spoil the answer to others.

Here is the plan made in free move mode, first step is 4s and 2s, since there isn't another/better way to play them.
Spoiler:


Then we can add the 3-chain and 6-chain.
Spoiler:


And finally add 1s, and rest of 2s, 3s, 4s and 6s, the empty spots are where 5s go.
Spoiler:


And the puzzle actually played according to the plan.
Spoiler:

Triplechain Quote
07-21-2012 , 08:00 PM
Quote:
Originally Posted by TH10
I should probably do what Gaby did, and give my puzzles some sort of difficulty ratings.
Done, 1-5, 1 is easy. Not sure if all are correct, wasn't that easy to compare them, but they shouldn't be off by more than one number.
Triplechain Quote
07-21-2012 , 10:34 PM
Just played about 50 Timed games against TH (around 2.5 hours). I forgot that Triplechain could be fun.
Triplechain Quote
07-23-2012 , 04:50 AM
A second and a first can't even gain me 1 rating point

reset all ratings imo
Triplechain Quote
07-23-2012 , 06:48 AM
Quote:
Originally Posted by ibavly
A second and a first can't even gain me 1 rating point

reset all ratings imo
Are you sure? That's surprising, I know the system is bad but it's not THAT bad is it? Your plays over the last 2 days correspond to roughly 20.5 wins, which should be enough for even me to gain one full rating point...
Triplechain Quote
07-23-2012 , 07:22 AM
Given the continual annoyance of finding yourself first after a few players have played only to find at the end of the day that you're 6th or 7th, I decided to work out the probability of winning if you're first after k players have played and n players play the game that day (modelling players as uniform r.v.s). I expected the calculation to be long and arduous. Turns out, unless I've been very stupid, that it's actually really simple: k/n. I worked it out using (very basic) combinatorics but now I see the answer I realise it's just obvious; the remaining (n-k) players always have a chance of (1/n) to win; you absorb the rest of the probability.

Sorry for spamming the thread with maths, I just quite like the simplicity of the answer and thought other people might be curious to know this.
Triplechain Quote
07-23-2012 , 07:27 AM
I also recently figured some math out. I knew if you play a timed match against an equally skilled opponent you have 50%.

Turns out if you play against 2 equally skilled opponents you only have 33%. lol

Last edited by ibavly; 07-23-2012 at 07:27 AM. Reason: You forgot to factor ties
Triplechain Quote
07-23-2012 , 08:13 AM
It was an approximation. Just like everyone being of equal skill was an approximation. We don't need an exact answer. And just cos an answer is simple that doesn't mean it's simple to arrive at. Not claiming this was hard or anything, but it did take me ten minutes given that I tried using Bayes' Theorem first.
Triplechain Quote
07-23-2012 , 08:18 AM
Quote:
Originally Posted by Paul101
Given the continual annoyance of finding yourself first after a few players have played only to find at the end of the day that you're 6th or 7th, I decided to work out the probability of winning if you're first after k players have played and n players play the game that day (modelling players as uniform r.v.s). I expected the calculation to be long and arduous. Turns out, unless I've been very stupid, that it's actually really simple: k/n. I worked it out using (very basic) combinatorics but now I see the answer I realise it's just obvious; the remaining (n-k) players always have a chance of (1/n) to win; you absorb the rest of the probability.

Sorry for spamming the thread with maths, I just quite like the simplicity of the answer and thought other people might be curious to know this.
is that really right, it sounds pretty sketchy to me after 15s thought.
Quote:
Originally Posted by Paul101
If I'm gonna make a play for a win, it'll be in the setup rounds. Like yesterday. I think my Round 4 play was probably the maximum score EV, but compared to the other play has a lower expected head-to-head winrate but a higher overall winrate. So there you go, that's me sacrificing rating for win EV.
i've said before, but i'm pretty sure that in the DC playing to max your score EV is wrong whether you are chasing rating or wins. If you're chasing ratings you want to max expected number of H2H victories, and if you're going for wins, you want to max prob(you win) (ldo). Neither of these things is trivially the same as score EV.

I've tried before to find a good example of where the first two differ, but it's not immediately trivial (i guess also because it depends on what other people are likely to do), so maybe for all intents and purposes it doesn't matter.
Triplechain Quote
07-23-2012 , 08:18 AM
also: first day back from a break, ship the DC. Second day, normal service resumed.
Triplechain Quote
07-23-2012 , 11:10 AM
Quote:
Originally Posted by kokiri
is that really right, it sounds pretty sketchy to me after 15s thought.
I'm fairly sure it's right... my guess is that you think it should be higher? Or not linear?

Quote:
Originally Posted by kokiri
i've said before, but i'm pretty sure that in the DC playing to max your score EV is wrong whether you are chasing rating or wins. If you're chasing ratings you want to max expected number of H2H victories, and if you're going for wins, you want to max prob(you win) (ldo). Neither of these things is trivially the same as score EV.
I agree. The quote you posted also agrees with this sentiment.

Quote:
Originally Posted by kokiri
I've tried before to find a good example of where the first two differ, but it's not immediately trivial (i guess also because it depends on what other people are likely to do), so maybe for all intents and purposes it doesn't matter.
When you say the first two, you mean score EV and head-to-head?
Triplechain Quote
07-23-2012 , 11:28 AM
Quote:
Originally Posted by ibavly
A second and a first can't even gain me 1 rating point
This isn't exact since I didn't know joet's rating, used the lowest rating any other player had instead, and the ratings of players may have changed by one or two, it should still be very, very close to reality.



Oh, and lol at the DC yesterday, more than half the players won!
Triplechain Quote
07-27-2012 , 09:18 AM
Quote:
Originally Posted by Paul101

Re-revised targets, now I believe I have a shot against TH again:

Combined Rating:

Short-term: Maintain the second-highest combined rating, and get to within 75 of TH.
Medium-term: Get to within 50 of TH.
Long-term: Beat TH.

Timed:

Short-term: Maintain 2nd and Master status in the Timed ranking.
Medium-term: Get to 1610
Long-term: Reach 1625 and pass TH (these will happen around the same time).

Daily Challenge Ranking:

Short-term: Get ahead of Dynasty.
Medium-term: Reach 1480 rating.
Long-term: Beat TH and get Expert status.

Daily Challenge Wins:

Maintain the highest Daily winrate, keep it above 1 in 9 and hopefully get it above 1 in 8.
Long-term: Top the win-list. This is the most-achievable long-term goal, although it's gonna take ages to catch Dynasty; around 3 years at the current rate!

Overall:

Break away from Dynasty and Gaby to become the number 2.
Long-term: Become number 1.
BOOM! Finally made it to within 75 points of TH. Only took 2.5 months to meet my short-term target :-/ New Targets:

Combined Rating:

Stay within 75 of TH.
Medium-term: Get to within 50 of TH.
Long-term: Beat TH.

Timed:

Maintain 2nd and Master status in the Timed ranking.
Short-term: Get to 1610
Medium-term: Get to 1620
Long-term: Reach 1630 and pass TH.

Daily Challenge Ranking:

Stay 2nd.
Medium-term: Reach 1485 rating.
Long-term: Beat TH and get Expert status.

Daily Challenge Wins:

Maintain the highest Daily winrate and keep it above 1 in 9 (I now think 1 in 8 is not incredibly realistic).
Long-term: Top the win-list. Preferably before Gaby does. If anyone is still playing by the time this happens.

Overall:

Maintain number 2 status.
Long-term: Become number 1.

Right now I'd say my most realistic long-term goal is definitely the Daily ranking one. It's also the one I'd most like to get. Long-term Combined and Timed ratings (beating TH) are now (much) less important to me because of the immobility of the Timed rating system and the infrequency of games. I'd really like Timed to pick up again.

Last edited by Paul101; 07-27-2012 at 09:47 AM.
Triplechain Quote
07-27-2012 , 10:40 AM
Quote:
Originally Posted by Paul101
Daily Challenge Ranking:

Stay 2nd.
Medium-term: Reach 1485 rating.
Long-term: Beat TH and get Expert status.
Really!? REALLY!? REALLY!? doublefacepalm.jpg

Quote:
Originally Posted by Paul101
I'd really like Timed to pick up again.
On that note. 'Are you tired of getting your aces cracked in Sunday donkaments? Or worse, can't even play them. Worry no more! Scheduled Sunday Triplechain Timed games are here!'

So in an attempt to revive Timed mode, let's try something new. Setting up weekly times to play Timed, hopefully this way we can get multiple people online at the same time. Let's start with one scheduled time though, because of people living all around the World, European weekend evenings are probably the best option. The first time to try is Sundays at 10:00-10:30 PM BST. Idea is to play weekly for 30 minutes (can obviously continue after that too) starting at 10 PM BST, first one is this Sunday. I'm obviously planning to be there, Paul has said he should be able to make it, hopefully some others will join too, everyone is welcome! Of course there is no need to participate every week, anyone can play when they want/are able to, also no need to tell beforehand if you are coming to play or not, can just show up. Although for the first week it might be a good idea to post itt if you plan to show up to gather some momentum.
Triplechain Quote
07-27-2012 , 10:43 AM
Posts like these are why I'm so hesitant to touch the ranking system. That said, I'm about to touch the ranking system.

Here are the changes I'm preparing to make. If you have gripes, air them quickly!

Daily: Increase min deviation to 10 (currently 5). Decrease deviation penalty to 10 (currently 20). Increase deviation by 0.033 points per day of inactivity, so approx 1 point per month (currently no increase).

This should have the following effects:
- ratings for active players already at min deviation (most of us) will stay roughly the same.
- people who have done better in the last few months will see a jump in their ratings (opposite for people who have done poorly)
- ratings for players who have been inactive for 1 year+ should drop out of the top 20.

I'm going to do the same for timed, although there will be no inactivity penalty unless timed actually ran on a given day (>30 games). Min deviation will increase to 10 there as well, while the multiplier will be reduced to 5. (currently they're at 5 and 10).

Problems I see (but am prepared to live with)
- ratings will change, possibly by quite a bit for some people
- TH10 will probably lose his purple rating (and more generally, people who still play other game modes and are likely to be as good as their rating indicates will still be penalized for DC inactivity). Not necessarily a bad thing, but not everyone will like it.
- may be bugs, may slow down the site while rankings are rerunning, standard stuff
- people who want to play timed but aren't awake when games run may get a (very small) hit to their rating sometimes.

I've had enough complaints that the rankings change too slowly that I think it's time to do something about it in spite of the drawbacks.
Triplechain Quote
07-27-2012 , 10:54 AM
Also, I have too many projects on my plate right now, so I'm making a push to get TC to a point where I don't have to touch it for a while. This is likely to be the last chance in a while to make requests, so if you have an idea, please throw it out. If you've already suggested something in this thread, don't worry about repeating it - I'll go through the last few hundred posts and make sure I look at all of them again.

I'll give #1 priority (by far) to changes that have the chance to bring in and keep new players. So many of the things I would like to do with the game rely on a bigger player base. There are only so many stats I can add, and I think we're already at the point where the benefit of new ones is pretty small.

I would love to create some new game modes that increase the variety and the challenge (new dice, bonus squares, different game boards, advanced mode style abilities), but there's not much justification for it when we have 20-30 active players per day.
Triplechain Quote
Triplechain
$25m Guaranteed WPM on CoinPoker
Join the action now
Daily Rewards • Splash Pots • CoinRaces
Triplechain

      
m