Showing posts with label Math. Show all posts
Showing posts with label Math. Show all posts

Thursday, June 7, 2012

Somer Blink: Utility Calculations

Disclaimer: In this post I'm going to use **MATH.** I don't mean to scare anyone away, but its something I wanted to delve into. Also, please correct me if I'm wrong on anything here.

 

At this point, most people have heard of Somer Blink. If not, its a lottery system using all in-game isk and it is run by a third party website. I've been intrigued by it for a while, and despite not being much of a gambler, I decided to give it a shot today. I threw in 250M isk to start, and figured if I lose it all, no big deal. I decided I wanted to calculate how many tickets I should buy. I started working on the math and then promptly got caught up in just blinking as much as possible. Out of the 250M I used to start, I got as low as 1.5M isk remaining. Then I won a few big ticket ships and got right back in the game. I'm currently sitting at 493M isk. At that point, I knew I was getting pretty lucky, so I wanted to actually run down the numbers for this lottery. The results surprised me.

To me, this looks like a classic problem in Expected Utility, which is a concept out of Economics and Game Theory. To be honest, I studied engineering, not economics, so forgive me if I've botched something along the way. Based on what I saw though, this is the formula for utility:

L = pA + (1-p)B

where L is the lottery's utility, p is the probability of option A, and 1-p is the probability of option B. To translate this to Somer Blink terms, the probability, p, is the number of tickets purchased divided by the number of tickets available. A is represented by the value of the ship minus the cost of the tickets purchased, and B is simply zero, since losing costs nothing (outside of ticket cost, which is already accounted for). So, our formula for Somer Blink is as follows:

L = (n/n_max)*A - n*bid

 By changing values of n (number of tickets bought) up to n_max (total tickets available for specific lottery), you can see the utility of each possible number bought. If we rearrange this equation, we get a very useful form for the utility. 

L = n * (A/n_max  - bid)

Very clearly, the utility is proportional to the number of bids. The question is whether or not the other quantity is positive or negative. As is almost certainly by design, I have yet to find a blink that has a positive quantity for the (A/n_max  - bid)  term. This means that utility is decreasing for every bid you place on an item. It also means that utility is negative for all n, indicating that there is no rational reason for you to bid ever
Taking a step back, this all kind of makes sense. If the lottery were set up so that you pay one eighth of the price of the ship, and there are eight slots, you have perfect odds. You should make your money back exactly (and by exactly, I mean statistically over time). Since this is a system where the bids are more than the respective share of the ship, you're supposed to lose over time. That's how Somer Blink makes isk.

There's one last important thing to note. You can buy enough tickets so that you automatically lose isk. Divide A by the bid price and round down to the nearest whole number. Don't be a dummy bidding to automatically lose money. 

What does this all mean? Probably not a lot at the end of the day. If you buy multiple tickets, you'll earn less when you win, but you stand a better chance at winning each individual blink. Like almost all lotteries, its not really what you'd call a good investment. On the other hand, someone still has to win, right? If you get lucky, then good for you and enjoy the ride. I've doubled my money already, why can't you? Best of luck.

Monday, March 12, 2012

Thorough Titan Tracking Analysis

Edit: After some input from Two Step, which you can see in the comments, I've decided to update the analysis with the most relevant possible data. Initially, I used 6 T2 Tracking Computers to keep on the same page as Kirith, but the updated figures show 3 Cormack's Tracking Computers, 3 Tobias' Tracking Enhancers, a Strong Drop booster, and the strongest implant available (+6% tracking). This gives a tracking of .02572, per Two Step.

Recently there's been some buzz about Titan tracking. In fact, one CCP dev (the name escapes me at this point) explicitly said that they have a team working on it right now. That's all well and good, but Kirith Kodachi's post on the matter inspired me. I decided to check it out for myself and share the results with everyone.

The problem people have right now is that Titans are blapping sub-capital ships with their tracking fit XL guns. I can understand that. Its no fun to be one-shot by something you're supposed to be able to out maneuver. Let's start with Kirith's first example: A Titan with 6 Ion Siege Blaster Cannons and 6 T2 Tracking Computers loaded with tracking scripts tries to take out an Afterburner fit Thorax. The cruiser has a max speed of 542 m/s and a signature radius of 140 m. I tried to pick ranges illustrative of the Titan's effective range of fire, so I chose its optimal (19km), its optimal + falloff (34km), and a point in the middle (25km). I think the only major difference in my analysis versus Kirith's is how we went about calculating transversal speed. My formula was this:

Transversal Speed = Speed * sin(Transversal Angle)

So my assumption was that the ship was going full speed and varied the transversal angle by manual piloting. Here's a quick diagram for anyone who isn't on the same page yet. I'll place a lot of emphasis on transversal angle as its a fairly intuitive concept in manual piloting once you understand what I'm getting at.


Also, here's the equation that Kirith found regarding chance to hit (from Evelopedia)



So anyway, here's how the results came out for the initial AB fit Thorax versus the Titan when considering Transversal Angle. A quick disclaimer: like Kirith, I'm not taking into account any leadership bonuses, remote effects or anything other than the basic Titan I described above. All of those things will make what you are about to see even worse (the little lines will move higher on the page).



So the results are pretty much what we expected. It essentially boils down to the combined effects of range and transversal speed/angle. The actual numbers here may be surprising though. The chance to hit is still below 10% for even just 20 degrees of transversal angle maintained, and at 30 degrees of transversal, even the farthest range still has less than 1% chance to hit. To me, this gives you one quick and easy rule for Titans: Dont Use the Approach Button! If you aren't flying right at (or away from) a Titan, its pretty darn hard to hit you. 

Let's consider that same Thorax with an MWD fit though. The only numbers changed are the signature radius and the maximum speed. What happens here?



Might be some problems with this one. Now even at 40 degrees of transversal movement, we've got 1/10 chance to hit on all but the optimal setting. The Titan has a 25% chance of hitting you out to almost 30 degrees of transversal. This engagement profile looks a lot more friendly for the Titan. If you aren't consciously trying to maintain transversal, it looks like you're in trouble. 

The situation gets even worse for a Drake. Take a look at this Drake fit including an MWD and shield mods and rigs that increase Signature Radius to 2188 m.



If that figure doesn't scream "Bad News" to you, I don't know what does. This basically says that if you're within an Erebus's falloff range, you're getting hit. Even at 90 degrees of transversal (which is full circular orbitting at that point, ie no "approach" or radial movement) you're getting nailed 40% of the time. Oh and by the way, you're getting nailed with an alpha of over 40k EFT damage. Enjoy that, Drake pilot.

The last angle consideration I made was the Tornado. The Tier 3 BCs were meant to be quick, to have small-ish signature radii, and to pack a punch. Maybe a Tornado fleet could have more encouraging results.



Even with the MWD on (speed 1782 m/s, signature 1020 m), we're back to seeing a nicer picture of things. Tornadoes can be pretty effective versus Titan tracking, but they had better still keep their transversal up. 

Summary
Graphs, EFT, and the Ships Above
I could turn out literally hundreds more graphs on the subject. If you're curious about actual DPS output versus range and transversal, hit up EFT. They've already got the software to give you pretty graphs and answer your specific questions. Everyone knows its not how the game really works because there are always other variables involved, but its also pretty plain to see a Titan can one or two shot pretty much every sub cap. If you've got any more questions on the hit chance mechanics, I'd be happy to provide any other graphs on the subject relating to any of the relevant variables.

In all honesty though, I'm not sure Titan tracking is totally broken. Both variants of the Thorax discussed, as well as the Tornado, can get under the guns of the Erebus without too much trouble. It might be a pain in the ass and take a while with some manual piloting, but its do-able. And in the case of the Drake, you can easily orbit outside of Antimatter's effective range (optimal + 2x falloff = 49 km) and sling missiles at it all day long. Now if he changes to Iron XL, that's a whole new ball game. 

The Glass Cannon
An important thing to take into account is that there will almost certainly never be a Thorax that takes down a Titan solo. The idea of a "glass cannon" with a few million EHP is a joke. They aren't glass cannons, so we can stop with that nick name. There should/will be support fleets on either side. If no one shows up to help the Titan, you can totally grab him like Jester pointed out not too long ago. Even if the Titan in this case was ruining the attacking fleet, you better believe people were reshipping. But, with a lock time of close to 30 seconds on an AB cruiser (before SeBos), he wasn't just running them off the field with his XL guns.

The Final Word
My final word is that I'm not the one to make the call. There are many who are infinitely more experienced with Titan warfare on both the attacking and defending sides of the issue. Personally though, I'm okay with the most expensive ship in the game being a bit on the powerful side. I've heard Titans quoted at 80 Billion isk before fittings. I'm more than fine with 80 billion isk going a long way in terms of DPS. Hell, if you could find enough pilots to fly them, 80 billion worth of T2 fit Rifters is going to pack a punch too.
Regardless, Titans do seem to be able to pop smaller ships pretty easily. This whole analysis ignored scrams and webs. Webs in particular are effective because they slow down the target without turning off the MWD and related signature bloom. The equation, which Kirith showed in his post, is very much dependent upon speed and signature radius. If a Titan pilot has a well coordinated support fleet, in particular with Rapiers/Huginns/Lokis, I could easily see how people aren't happy about Titans tearing them apart once they're webbed. And multiple Titans at different ranges? Yeah. Something doesn't seem to be stacking too well here. 

In the end, its all situational -- just like everything else in Eve. If its one Titan, no corresponding support fleet, and you've got enough AHACs on board then go for it. On the other hand, if its multiple Titans and a well organized support fleet to slow & hold your guys down? Well... good luck. I'm not inherently opposed to some rebalancing, but I wouldn't know where to start except for maybe increasing the signature resolution for XL guns (or at least Titan XL's). I'm looking forward to what CCP puts out in terms of the impending Titan changes.