Topic: [ANN][MOTO] Motocoin - page 74. (Read 178259 times)

Give me your map preselection algorithm and I will go through the map manually. Pressing F6 and visually analyzing map is boring and time consuming process. I was never that lucky as your algorithm according to the maps you provided. How many maps your preselection algorithm is reviewing in one second? It is impossible to accomplish the same manually. Bigger map will broke your preselection phase. You wouldn't be able to enumerate that many maps to find one without the obstacle on the way from start to finish in time.

Bigger map will also make much more difficult to decide where to go. The heuristic will be more difficult for a bot creator to code.

Also we should take into account that you are a bot owner and you aren't interested in arrangements against bots. You can try to foul the trail

If you do indeed achieve a bitfinex 2.0 for alt coins you'll have the absolute best crypto exchange for alt coin pairs...EVER.
I'd 1000% trade there!

The map generation itself would have to be changed rather drastically.  Again, I don't want to say too much and give away some of my edge, but perhaps the phrase "early bailout" means something for you?

Look closely at the map "seed" generation and how it relates to the perlin function...

Unless that relationship is changed dramatically, the bailout will save the bots, there.

But this is probably already too "telling" now, so I won't say more. 

Maybe UTC should be used as a nonce for map generation, I think it's safe to assume that 50+% motocoin users have correct system time (+/- 5 seconds)

Will it be a restart of the coin or just fix?

At least it will prevent bots from selection such a simple map. Those maps you provided are also very easy to go by a human not only by a bot. Your "pre-processing phase for map selection" shall be made impractical. The number of random obstacles on a way from start to finish shall be big enough to prevent you from choosing a map with a simple route through the simple iteration over few maps. I am really confused about this flaw in game design. I thought that the map is huge enough and prevents such a hack.

Thank You very much. Will take to serious consideration.

It is funny you should mention this.  Yesterday I was actually playing with some "try to always move toward coin" type heuristics and found that it actually *significantly* decreased production!


The F6 key is very important for both humans and bots.


You can do nothing to prevent them, only mitigate them.  On a long enough curve, bots will out-think us on anything we throw at them.

Welcome to the new age.

I, for one, welcome our new robot overlords.


Changes to map generation semantics will have less effect than you might intuitively think.  The best changes to map generation will be ones that make TargetTime more relevant.


Actually does practically nothing to hamper the bots, on its own.


This is kind of "the litecoin argument" I guess.  Scrypt has ASICs now, X11 will not be far behind.  You'd need to make map generation in-feasibly difficult for this to really "matter" - and it would significantly hamper human play in the process. (Imagine waiting 5-10 seconds each time you pressed F6?)


Here we get to the crux of the matter.  We are facing an asymmetry with each-other.  You will spend days/weeks implementing changes to deter the bots, and might even knock the more basic ones out of the running, but the good AIs will adapt within minutes/hours of your hard forks, in perpetuity.  This is not a bad thing!  THIS is why I say coins like these will offer direct incentive for AI research.

GAME ON!

I meant it as something of a joke, but yes.  I was not particularly aware of c-cex or the ways in which it is a unique offering in the space until MOTO.  Perhaps you should do more advertising targeted at potential market makers?


If done like Bitfinex ("peer lending" swaps model) it should work exactly the same regardless of if the swap is BTC/USD or DRK/LTC... you just keep a CFD pair with an attached interest schedule, and a cron job to process the interest payments on positions.


Although this is something often "rallied against" in the crypto-currency space having these sort of control policies is something that (traditionally, IRL) distinguishes "real markets" from fly-by-night operations.

Main reason why this was botted is because of these fall-thru maps. Bots probably just find these maps and then play them looking several seconds into the future and choose actions that will move them closer to the coin. I believe, it is much harder to complete usual map with bot because sometimes you need to actually move in opposite dirrection from the coin.

What can we do to prevent bots?
1. Change level generation algorithm to prevent fall-thru maps. This is the main countermeasure against bots.
2. Increase map size.
3. Use some sort of Scrypt/X11 in map generation to prevent fast map enumeration.

This will definitely take some time to implement.

Smart man, indeed.

I'll not say too much, as I wouldn't want to give away too much of my edge to the other botters.

However, yes, there is a pre-processing phase for map selection, and this is critical for anyone wanting to make even a naive bot be successful.  This is (relatively) time consuming, since the map generation is based on SHA512 rounds.  (My "best bot" actively tries to balance time spent on map selection and time spent on terrain traversal.)


The bots would just have to modify the map selection process some to compensate.  Basically, a larger map would deter the bots, but not in the way that you might think.  (Perhaps we could discuss this further in PM/IRC or something?)


Skilled players already do this map processioning step themselves, as well.  A good player will hit f6 in S mode quite a bit, quickly assessing maps, and then will zoom back in to fly to their coin.  (My whole bot design is modeled after observation of how a successful human player plays.)


True, and this would make the balance between "traversal time" and TargetTime threshold even more relevant, as well.... but it will not actually slow the map selection process much.

If you really want to "beat the bots" however, the solution would be something like random obstacle courses.  Right now, the fact that the game has 0 "modality" in the solution space is what makes it most weak to bots.  If completing a course required a selection of skills applied at different times (instead of "the same 2-3 skills" applied continuously through the whole run, as it is now) this would significantly hamper the bots.  (It would also make the coin an even stronger driver of AI research, hehe.)

Liquidity - You mean trade volumes? More traders?
Margin - yes, good and standard feature for fiat exchanges but very hard to implement for crypto.
We do market freeze only by request of coin owner/developer. We did not receive any request to freeze from MOTO.
Will learn Bitfinex deeper. Thank You.

More liquidity would be nice.  Margin would be nice.

Market freeze policy to insulate panic dumps like MOTO saw in the past 12 hours would be very nice, so we aren't left with order-of-magnitude spreads. 

Basically just become Bitfinex for altcoin pairs and I'll be very happy, myself.   

The maps are also looks rather simple. Is this a pure luck or your bot also iterate through the maps? How much CPU time your bot needs to decide if the map is easy to go or not? If this is easy then the map shall be enlarged to prevent such a hack. Or we will see a bots competition on an easy to go maps in a nearest future and human will have no chance there for sure.

Hm... actually map enlarging will also make mining process more interesting and less nervous. Skill will be more important on a bigger map then a pure luck.

Hello!
Being the only exchange for MOTO at the moment I would like to ask You for some feedback of our service.
Are You satisfied? Maybe something to improve? Any suggestions?
Thanks in advance.

THIS.  Smart man, you are.

It will be difficult for people to understand that this is the first coin where the difficulty target does not directly affect the block time with a single bound, but instead indirectly bounds it with both a high and low "cap."  (There is a hard "wall clock time" block limit in addition to the Target threshold, and the target has a lower bound on time! (as obviously no bot or human could ever grab a coin in just 10 frames - there is some minimum travel time required to the run.  Eventually the goal will not be just to solve levels, but will be all about minimizing the travel time!))

The ability for this coin to maintain a security threshold without just "burning energy relentlessly" is why I consider it the first true 3.0 coin.  Some other coins (incl HUC) came close, but their security is still backed primarily by energy spend.  (Or else they are fatally flawed and failed/failing otherwise, heh.)


This is the current state of things.  The bots are all "naive" and simple, for now.  Already it is unlikely a human will win without quite some skill (and hitting that "speed up time" button a lot) but the bots are still competitive with *skilled* humans at these TargetTime values.


This is a potentiality, but I think it is probably a long way off, still.  (I hope that I'm wrong!  The AI Winter has now thawed, after all!)

i thank it is a good game,but not a good coin~just i think

to be a honer , i still can not get even one ,it's so difficult for me .lol

I get "Corrupted block database detected. Do you want to rebuild the block database now?" every time I start my wallet... Does anyone else experience this problem?

Of course you can see all of my bots' wins on the blockchain itself.    (BTW I might soon be collecting the remaining bounty for the web viewer as well, heh)

Here are some of my favorite plays by the favorite of my various bots, which I call "flipper" for its tendency to flip around wildly:

#9147 - The first coin that this particular bot grabbed.  It is funny because there were still several bugs in the algorithm.  The bot used all 60 of its moves immediately, for example.
#10573 - Most bugs fixed by here, a cute run.
#10925 - The last play by the first main iteration of this bot's code.  You can see it is already much improved here, with most of the bugs in the basic implementation sorted out.
#11053 - The second major iteration of the code, and where the bot started really earning the flipper name.
#11061 - Just one of my favorite blocks, I like how the bot bounces off the ground upside down at the end to grab the coin.
#11164 - Some slight improvements to the acceleration control algorithm, now with even more tightly controlled flippage.  I basically "borrowed" some tricks from the other bots to optimize at this point.

From there this bot was basically left to run as it was, while I moved on to developing on "smarter approaches."  The replays after that point get tediously repetitive to watch.


Hah, by my own definition of "working" I still haven't yet! (I'd like the algorithms to be much smarter than they are, still!)  However my first botted coin was grabbed just a few hours after this thread was started.


Coded from scratch, more or less.  I'm using some off the shelf components for some "obvious" things (I wasn't going to code up yet-another-evolver when I could just grab FGA or Beagle, hehe) but the bots are entirely my own design.


Not to my knowledge.  Since I consider some of the technology involved "proprietary" I probably would not be inclined to sell it, either.  However, if anyone is really interested I suppose I would listen to offers.  (I have considered setting up something like a mining pool, where users could contribute resources to the bots' calculations in exchange for a share of the coin mined.)

Also, I will be benevolent with my mining power.  I have no intent to 51% attack or anything of that sort.