I picked up a big box of tomatoes from the farmer’s market. When I got home I went searching online for tomato sauce recipes, looking for a basic recipe to start with. My idea was to start simple, gradually adding variations and experiments as I worked my way through the box. To my surprise, there doesn’t seem to be a consensus on how to make a “basic” tomato sauce. The tomatoes can be plum, roma, beefsteak, or many others; red or yellow; fresh or canned; they can be peeled, seeded, chopped, crushed, or any combination of those; they can be roasted, sauteed, or simmered; the sauce can be cooked for anywhere between five minutes and 10 hours; it can be blended, passed through a ricer, or left as-is; it can be flavored with salt, pepper, onion, garlic, oregano, basil, chiles, and countless other herbs and spices. On top of that, the quality of the ingredients, the size and shape of your cooking pot, your stove, humidity, and elevation undoubtedly play some role in the final product.
I realized my plan of starting basic and proceeding by trial-and-error was in no way guaranteed to yield a great tomato sauce. There were too many variables. Even if I spent my whole life on tomato sauce I could never test every combination of ingredients and methods, and even if I could, the results would still be inconclusive. Every time I changed one thing, many other things would change as well (temperature, seasonality of tomatoes, etc.).
I decided to take a more theoretical approach based on general principles. I would use the big, round tomatoes I had — that part was easy. I wanted a lighter sauce, so I eschewed roasting the tomatoes. I didn’t seed or peel the tomatoes, because it’s a pain and seems a bit wasteful, and anyway I don’t mind the seeds and I actually like how the peels maintain a chunkier texture. I stuck to only the most basic, essential flavors. I softened onion and garlic in olive oil before adding the tomatoes, seasoned with salt, and cooked until the texture seemed good, maybe about two hours. Was it the best sauce I could have made? Probably not. But it was pretty good.
Awhile ago I referred to, “Who’s got the chips, bro?” as the fundamental theorem of poker. I was partly joking — there is a well-known Fundamental Theorem of Poker, put forward by David Sklansky, that says that any time your opponent plays differently than they would have if they could see your cards, you make money. But “Who’s got the chips, bro?” does have its advantages. Raw cash results are the most basic piece of evidence you have as a poker player and the most closely tied to your ultimate goal, winning money.
There are some situations where results are obviously misleading. If you get all-in and need running cards to win, even if you actually hit those cards, that’s a situation you should try to avoid in the future if you want to make money. We know there’s a lot of luck in poker. In other words, if what we’re after is good plays, there’s a lot of noise around the signal. That means much of the time we have to disregard short-term empirical results in favor of pre-existing theories. Nonetheless, it can be difficult and troubling to ignore results, especially given how many poker players rely on false or outmoded theories that make them blind to their own mistakes.
Ultimately there’s no easy answer. My natural inclination is to see theory as “just making stuff up” and empirical evidence as more reliable. But when dealing with complex problems with many variables, empirical evidence is also problematic. In online poker it can take upwards of 50,000 hands to develop an adequate sample size for some stats. Those 50,000 hands include hands against a huge variety of opponents, played when you were in various frames of mind. You can never play the same hand twice: table conditions have changed, and you have changed. Even results gleaned from large samples aren’t entirely trustworthy. Theory has the advantage of getting at some broad ideas or patterns that might take a long time to emerge from empirical data. It has the disadvantage that it’s easy to screw up and difficult to test.