I agree with you, using quantitative methods to do long term predictions is... a little bit tricky.
But we are not forced to do long term prediction. This is why when using quant models, the prediction should be made on shorter time frames and taking advantage of smaller market movements. The biggest great side effect of this would be an increased Sharpe ratio(at the expense of higher trading cost).
And regarding inefficient markets: I stick to my view. The more the price is above/below the right price, the more trading opportunities.
The thing is that there is no correct answer, some can make money using fundamentals, others using systematic algos while others use both. So getting back to your initial claim "fundamental information about the asset is much more important/valuable than mathematical models" is wrong because it depends on who you ask: a systematic or a fundamental/value trader? As I showed you above, the mathematical models can generate huge profits.
I just spent a while reading about this Medallion fund, and it looks very interesting. It looks too good to be true, which leads me to believe it is not true. A lot of people seem to speculate that Simon's returns may be false and they could be used to lure people into investing in his other funds. There is no proof of those returns anywhere and the fund is so secretive that there is no way to verify anything they claim.
Anyway, regardless of whether or not the fund's returns are real, I have a few thoughts on the issue:
1. The most successful investor of our time is Warren Buffett, and he is clearly all about the opposite of this - fundamentals beat mathematical analysis. Warren Buffett's success can be verified, whereas Medallion's cannot.
2. The smartest people in CS/AI/Math are not working at hedge funds - they are working either at a company like Google, or their own startups. People who are this good in those fields rarely care about money, and care much more about impacting the world with their research/technology. They mention that a lot of the researchers at Medallion were working on speech recognition and AI at IBM before going to Renaissance - if they were worth their salt, they would have stayed at IBM and made massive strides in AI, generating far more money than they ever could at a hedge fund. By the way, I say this as someone who has worked with many of the smartest people in these fields and who works in CS today.
3. Trading only based on mathematical patterns means that you ignore the real value of the asset, which can change based on things that aren't considered by mathematical models (such as analyst research reports, unfounded rumors, etc). This method will at least lose money sometimes. If a trade based on quant goes in the wrong direction, there is no assurance that it will come back and bring a profit, so you are likely better off exiting your position. With fundamentals, the story is very different. If you determine the intrinsic value of a company to be a certain amount and make a trade based on that information, even if the price moves in the opposite direction of your trade, you can simply double down and increase your profits when it does finally come back to its intrinsic value, which you know with a very high probability that it will. We know that the markets have enough efficiency that asset prices will eventually meet their intrinsic value, even if it takes a little while. Basically, my point is that your chances of losing money are greater if you don't consider fundamentals, whereas not considering technical/mathematical patterns does not mean you will lose money. Even if the trade goes south, you know it will come around eventually if you have done everything right in valuing the asset.
The more the price is above/below the right price, the more trading opportunities.I agree with this, but only because the markets are efficient enough to eventually correct this. The more inefficient they are, the less likely they can correct something like this, eliminating the opportunity.