Slashdot Mirror
Can Statistics Predict the Outcome of a War?
StatisticallyDeadGuy writes "A University of Georgia scientist has developed a statistical system that can, she claims, predict the outcome of wars with an accuracy of 80 percent. Her approach, applied retrospectively, says the US chance of victory in the first Gulf War was 93%, while the poor Soviets only had a 7% chance in Afghanistan (if only they'd known; failure maybe triggered the collapse of the USSR). As for the current Iraq conflict: the US started off with a 70% chance of a successful regime change, which was duly achieved — but extending the mission past this to support a weak government has dropped the probability of ultimate success to 26%. Full elaboration of the forecasting methodology is laid out in a new paper (subscription required — link goes to the abstract). Some details can be gleaned from her 2006 draft (PDF)."
Is that lots of people are going to suffer and die, and lots of money will be spent, usually with detrimental results to all parties involved.
Oh yeah, and the companies that make bombs and guns will get richer.
it's a blue bright blue Saturday hey hey
She took a bunch of historical information about wars, built a model and then when run on that historical information it was 80% accurate.
Amazing stuff.
Hari Seldon, is that you?
As far as I know nobody has formally specified the 'win' outcome for the war -- so I'm a bit doubtful that anyone has worked out an EXACT 26% (not 25%! That number would sound like a guess! But 26% sounds like SCIENCE!) chance of the US side achieving it.
If the 26% really was worked out with a reasonable methodology, then the interesting part isn't the number so much as whatever definition they came up with of 'victory'.
That said, giving ridiculously exact answers to impossibly vague questions is fun and harmless. 92.8% of the time.
Whence? Hence. Whither? Thither.
Sometimes others will decide that you are going to play it whether you want to or not.
TWW
"Encyclopedia" is to "Wikipedia" what "Library" is to "Some people at a bus stop"
The problem with this is that the model was "trained" on the same historical data on which it is eventually tested. This doesn't prove anything.
As an example, a defence contractor once built a system that would recognize wether or not a tank was in a picture. First the system was trained on half the "with tanks" and half the "without tanks" pictures. Next the system got a good percentage correct on the second half of the pictures. It turns out the "with tanks" pictures had been taken on a sunny day, and those without on a cloudy day. So the system was actually telling "sunny" or "cloudy".
In this case, it could very well be that her system predicts the outcome of the war, based on the weather in tokyo 6 weeks before the start of the war. This example was chosen so that you, not an expert in this field, immediately can dismiss this as a nonsense predictor. But as the model gets more complicated, and you feed it lots of parameters that might seem relevant, even the experts will no longer be able to see the value of such complicated predictions. At some point you just have to "trust the computer".
Aerodynamics: Yes. We understand the underlying principles, we've verfied the predictions made by the models in real life, and found that it matches very good.
In this case: No. Before I trust such a model, it would need to be verified (as is, no modifications allowed!) against say at least 20 wars that haven't started yet. If it preditcs the outcome of those correctly, the model has merit.
I'm not going to wait around (I hope).
I know I'm going to get modded offtopic at best here, but wasn't the point of the invasion to stop Iraq from deploying its extensive stockpiles of WMDs? Wasn't it supposed to be a pre-emptive strike to get him before he got America and its allies? Toppling Hussein was supposed to be a byproduct of that, but it was not the primary goal.
/because/ of the invasion? Then, erm... SUPPORT OUR TROOPS!
You're right, statistical prediction is unlikely to suceed if every couple of months the goal of your war changes, but its next to impossible if even the initial point of the war gets retconned. How far can this go?
We're going into Iraq to stop Saddam and his WMDs.
No WMDs found? Oh, then we came to Iraq to stop Saddam and free the Iraqi people.
Saddam gone and there's still fighting? Then we came to Iraq to fight the terrorists there so we dont have to fight them here.
Terrorism worldwide increasing despite, or possibly
How far can you push this? No statistical model, no battle plan can succeed if the people in charge can't even make up their mind what they are fighting for.
That said, I find this very unconvincing. And why? Because it is actually very hard to measure the outcome of a conflict, especially when the actual strategic objective of the conflict may be a state secret on the side of the aggressor. Put simply, we do not really know, in the case of Iraq, what the real objective of the US Government is. Is it:
- To stabilise Iraq with a government that will be a more reliable client of the US than Saddam was?
- To destabilise Iraq and the Middle East to prevent power accumulations that threaten the US regional aircraft carrier, USS Israel?
- To maintain high oil prices by creating instability, enriching the Bush family and their clients?
- To keep up pressure on other states by showing that the US will intervene and create anarchy if it wishes (The old "remember what happened to XXX country?" "There is no XXX country." "Exactly, that's what you need to remember")
My point is that for any desired outcome other than the first, the US Government would be achieving its strategic objectives. The fact that these objectives might be objectionable to the majority of the US population is irrelevant; for most of history, wars have been fought by military elites without reference to the interests of the majority of the population. In exactly this way, Vietnam can actually be seen as a victory for the US if the strategic objective was to stop the expansion of Communism. Personally I don't believe in the domino theory, but if you do you can argue that the example of Vietnam stopped other regional states from going Communist.In the past, wars usually ended when one side ran out of resources, whether provisioning, human, strategic or geographical. The constraint on warmongers in democratic societies is that society can ultimately strangle the resources of its internal warmongers without, necessarily, killing anybody. It is also possible for democratic societies to change the playing field so that strategic objectives change or become irrelevant. (e.g. by doing so much business with other countries that it becomes impossible to pursue strategic objectives without doing more harm to yourself - which you could say is happening with the US and China.)
Pining for the fjords
How many people have walked past stonehenge since it was abandoned without realizing it can be used as an astronomical clock? When the roman empire collapsed it took Europe 1000+yrs to relearn plumbing ( some parts of the UK are still catching up :).
It's not the equipment that's irreplaceable, it's the people.
And did you exchange a walk on part in the war for a lead role in a cage? - Pink Floyd.