Recommend
2 
 Thumb up
 Hide
61 Posts
Prev «  1 , 2 , 3  | 

BoardGameGeek» Forums » Gaming Related » Trades

Subject: TradeResolver version 8 out! rss

Your Tags: Add tags
Popular Tags: Math-Trade-Info [+] Math_Trade_Discussion [+] [View All]
Mikko Saari
Finland
Tampere
flag msg tools
http://www.lautapeliopas.fi/ - the best Finnish board game resource!
badge
Avatar
mbmbmbmbmb
Re: TradeResolver version 7 out!
mlvanbie wrote:
I could decide that any result that doesn't have me trading my games isn't worth sending in, or anything that doesn't have me receiving a particular game.


But it's a choice of sending in all the results or sending in no results at all. You can't make the decision to send in the results after seeing them - you only see them after they've either sent or will never be sent. And whatever results people present, the server will only take the solution that's the best so far.

Well, of course, since the server architecture is public and pretty simple, it's fairly easy to hack, but of course, blatant hacking is obvious and subtle hacking will provide less benefits. And the day someone actually goes that low for few bloody metal cd's, I'll quit hosting trades, period.

Besides, these doubts are theoretical - in practise the system worked really well, many people were interested in helping (and that even includes one non-participant) and most were even happy with the results.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
James Perry
United States
Spring Hill
Tennessee
flag msg tools
badge
Ex-Term-In-Ate
Avatar
mbmbmbmbmb
Re: TradeResolver version 7 out!
mlvanbie wrote:
You might think that the actions of one individual are unimportant, but I have access to more CPU power than you are likely to get from all of the other participants combined.

That is quite a bold statement given that a large majority of gamers are computer people. In my house alone I have eight computers when you add in work where I could concievable run this program on every computer there as a task after the company closes for the day. That is quite a bit of computing power from me alone. And no I don't think that I have even remotely close to as much available computing power as some people.

mlvanbie wrote:
I am not saying that there is anything wrong with getting help from self-interested sources, just that the trust model should be appropriately stated. Getting the trades that you want by presenting selected results is a game, and all of the traders are playing that game as competition. If you assume that everyone is gaming the system, then assumptions about finding best trade sets don't apply. It becomes a game of chicken in which people may have several results of varying quality (trade count, happiness, or whatever) with varying self-benefit. They choose the result that is best for themselves without being so bad that someone else will submit a better solution. How bad the submitted solutions are is a matter of groupthink and whether there exist solutions that are both globally happy and desired by single individuals.


Also remember that there are at least four people here on the geek that run these math trades on a regular basis that do not game the system at all. The four of us will likely be involved in this as well plus many other HONEST traders. Some people will try to game the system, and their results will be summarally crushed by size and/or quality of those honest participants. Eventually those that game the system will give up trying or be detected.

Further once Mikko gets the system fully functional, then as the trade program runs it automatically turns in its next best result set. Thus minimizing any chance of gaming the system. The only solution is then to hack the system.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Michael Van Biesbrouck
Canada
St Catharines
Ontario
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
Re: TradeResolver version 7 out!
Regai wrote:
mlvanbie wrote:
You might think that the actions of one individual are unimportant, but I have access to more CPU power than you are likely to get from all of the other participants combined.

That is quite a bold statement given that a large majority of gamers are computer people. In my house alone I have eight computers when you add in work where I could concievable run this program on every computer there as a task after the company closes for the day. That is quite a bit of computing power from me alone. And no I don't think that I have even remotely close to as much available computing power as some people.


I regularly submit jobs to 150 processors, some of which can run multiple jobs simultaneously. I have access to hundreds more that I don't both to use. At one point I did have 120 copies of TradeGenie running simultaneously. But the point isn't how much CPU power I have access to, it is that some have more than others. And many people will not use all of the power at their disposal.

Regai wrote:

Also remember that there are at least four people here on the geek that run these math trades on a regular basis that do not game the system at all. The four of us will likely be involved in this as well plus many other HONEST traders. Some people will try to game the system, and their results will be summarally crushed by size and/or quality of those honest participants. Eventually those that game the system will give up trying or be detected.


I wasn't writing about the people that run the system. I was writing because Mikko repeatedly made falsifiable claims about alleged security improvements made to TradeResolver. Mikko has made a great program, but he shouldn't be telling people that it does stuff that it doesn't.

Anyone successfully gaming the system is sending in better results than anyone else, or failing to send in a better one. Overall, it is detectable but it is not clear to me that Mikko was going to add in the extra code to do so.

Regai wrote:

Further once Mikko gets the system fully functional, then as the trade program runs it automatically turns in its next best result set. Thus minimizing any chance of gaming the system. The only solution is then to hack the system.


Yes, I implied that it was a trivial change to fix this.

Also, the essential security feature assumes that someone will check the results not just of the winning solution, but those of everyone else.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mikko Saari
Finland
Tampere
flag msg tools
http://www.lautapeliopas.fi/ - the best Finnish board game resource!
badge
Avatar
mbmbmbmbmb
Re: TradeResolver version 7 out!
Really, I don't see the problem here.

Case A: Let's assume you run TradeResolver and find a solution that's very good for you. The program sends it in and you quit the run to prevent the program from finding other solutions, that might be worse for you.

Case A1: The solution is the best anybody finds. It is used, and everybody's happy to get lots of trades. Edit: if you also find an even better solution, but somehow keep it to yourself, then sure, that's somewhat dishonest, but it's also something that simply can't be controlled.

Case A2: Someone finds a better solution. That solution is used instead of your solution. Everybody's happy to get lots of trades.

Case B: Let's assume your run of TradeResolver finds a good solution that isn't good for you and you have the skills to see that in advance (that's possible, but takes lots of effort) and you somehow manage to stop TradeResolver from sending in the results to the server (which, I do admit, happens right after it has written the results to a local file). A different solution is used and since somebody else (probably few people) is running TradeResolver for several hours, it's probably a pretty good solution. Everybody's happy to get lots of trades.

Ok, now it's your turn. Present case C that is somehow problematic.

The results will include a log, that contains the random seed that produced the results. Somebody can run the program with the same seed to see if it the correct results will come up. If they don't, then something's wrong. Whether or not running TradeResolver longer will come up with a different, even better set doesn't mean anything - there are several legitimate reasons to stop the program whenever. After all, I just asked people to run it as long as they can.

Also, I still can't see how you can cheat the check. If I run the program with the same random seed, I get the same results. If I don't get your results, you've cheated. Or is there a way to fool that check?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Michael Van Biesbrouck
Canada
St Catharines
Ontario
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
Re: TradeResolver version 7 out!
I agree that A and B aren't that bad. They are also not what you have claimed, violating the proposed model. I have repeated outlined A and B; it took a long time to get this far. Now, if people can cherry-pick the results that are produced, why do you care that the results they suggest correspond to a valid execution from a particular seed? Does anything matter other than that you select the best of the results sent to you?

Case C may be, what if everyone does it? Then the results are not as high-quality.

An interesting (but not necessarily bad) case D has everyone providing results by their methods of choice.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mikko Saari
Finland
Tampere
flag msg tools
http://www.lautapeliopas.fi/ - the best Finnish board game resource!
badge
Avatar
mbmbmbmbmb
Re: TradeResolver version 7 out!
I don't care - I didn't check the best results in our metal trade, even though it wasn't my computer that found them. I had no reason to suspect there's any cheating going on (besides, it would've taken many hours to check them).

That's one thing: you're already trusting your fellow traders a lot, so why not trust them here?

Well, anyway, the random seed option wasn't added for security. It was added because Joe Huber wanted it, and he wanted it for repeating experiments, I think. It was just an afterthought that it can be used to verify results. I still think it can.

Quote:
Case C may be, what if everyone does it? Then the results are not as high-quality.


I assume you're talking about my case B here... well, yes, if everyone does it, the results aren't as good, but in an environment like that, I wouldn't want to host math trades anyway.

As it is, it takes effort to cheat, since TradeResolver sends the results automatically. You have to be a bit of a hacker - anybody's clever enough with computers to run TradeResolver given a complete package and simple instructions, but gaming the system and preventing it from sending in certain results is not obvious - it takes some thinking, and it's clearly bending the rules.

Thus, I'm fairly certain that with most trades, there are just few people who even have the idea that it could be possible to cheat and those few people aren't probably going to do it, because they're honest people. If they're not, them manipulating the TradeResolver results is least of your worries.

Quote:
An interesting (but not necessarily bad) case D has everyone providing results by their methods of choice.


That's not a bad idea. Just forget any quality measures and count the number of trades. I could try that.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mikko Saari
Finland
Tampere
flag msg tools
http://www.lautapeliopas.fi/ - the best Finnish board game resource!
badge
Avatar
mbmbmbmbmb
Release 8 is out. It features distributed solving and duplicate checking.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Jeff Michaud
United States
Longwood
Florida
flag msg tools
On-Line Want List Generator - Hopefully Making Math Trades a Little Bit Easier
badge
Captain Kirk, Captain Picard, Captain Sisko, Captain Janeway, Captain Archer
Avatar
mbmbmbmbmb
GNU Java to native code compiler
Curious if anyone has tried using GNU's Java compiler which can compile to native (vs. java byte) code, on TradeResolver?

It should give TradeResolver a speed improvement in theory...

See http://gcc.gnu.org/java/

Jeff
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Jeff Michaud
United States
Longwood
Florida
flag msg tools
On-Line Want List Generator - Hopefully Making Math Trades a Little Bit Easier
badge
Captain Kirk, Captain Picard, Captain Sisko, Captain Janeway, Captain Archer
Avatar
mbmbmbmbmb
For what it's worth, I've since installed Cygwin (GNU tools environment under Windows), which includes the compilers.

Was pretty easy to compile TradeResolver with gcj and have a binary. The binary does run, it seems to read the properties file just fine, but then has some kind of java io error (invalid argument) when it tries to read the wants file.

Haven't had a chance to look past that... so have no data on how much of a speed up (if any) it may provide.

JeffyJeff wrote:
Curious if anyone has tried using GNU's Java compiler which can compile to native (vs. java byte) code, on TradeResolver?

It should give TradeResolver a speed improvement in theory...

See http://gcc.gnu.org/java/

Jeff
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
I'm free-basing midichloriens and the force is, like, an energy that connects and penetrates us all, man.
United States
Saint Ann
Missouri
flag msg tools
badge
Avatar
mbmb
msaari wrote:
Release 8 is out. It features distributed solving and duplicate checking.

Has anyone tried this out? Distributed solving sounds snazzy: how well does it function in practice?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mikko Saari
Finland
Tampere
flag msg tools
http://www.lautapeliopas.fi/ - the best Finnish board game resource!
badge
Avatar
mbmbmbmbmb
dreadpirate wrote:
msaari wrote:
Release 8 is out. It features distributed solving and duplicate checking.

Has anyone tried this out? Distributed solving sounds snazzy: how well does it function in practice?


I've done it several times, and it works great. How effective it is depends on the number of folks you can get doing it, but running the client is very simple, once someone sets it up. It can easily multiply the computing power.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Prev «  1 , 2 , 3  | 
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.