Investment thinking in the process of polishing short-term quantitative models

In the past two years, I have been mentioned from time to time as a long-term lazy person who is playing with a short-term quantitative model. Although the backtest data in the past 5 years is very beautiful, I still only use a small amount of funds for real-time debugging and optimization, and I dare not invest more money rashly. , in addition to being out of awe of the market, there are also some practical concerns, among which the investment thinking is sorted out and shared, hoping to get others to be elegant.

* As stated before, historical backtesting is only a very limited validation of the validity of an investment model. There are endless possibilities for the future we invest in. Even if you think that the model “follows” certain market principles, you still need to be cautious before finding enough systemic risk points to guard against.

* In order to make the model have a certain adaptive balance, I added a hard measure to limit the worst case: underperforming the CSI 300 by more than 5 points in a single year. More than ten years of back-test data show that the overall return is still good, but there are always a year or two that fail to meet the standard, and all of them occur in rising or even bullish years. Of course, the weak bear market and the bull-bear turning year (such as 15 years) can significantly outperform the broader market. No matter how you adjust the model stock selection or take profit and stop loss settings, it has not reached the target for the time being. Am I putting too much emphasis on risk control? (Even if the bull market relaxes the settings, it is necessary to prevent risks) Or is it impossible to have both?

* The bull market may underperform the broader market by a lot, and one of my original goals for polishing this short-term model has not yet been achieved. Although the price investment model I use now is not bad, it is relatively conservative. Especially in the late stage of the big bull market, it is basically only selling but not buying, and there is a high probability that it will underperform a lot of aunts (of course, after the bull-bear transition, I don’t know how to stop those Aunt will lose miserably). Polishing the short-term quantitative model and moderately participating in the opportunity in the later stage of the bull market, I want to make up for the shortcomings of the existing model. (The existing model will not conduct relatively short-term value speculation during the bull market bubble period, because it may not be possible to buy it)

* Accidental factors may cause a large drift in single-year returns. The priority of stock selection, the starting point of backtesting, the contingency of individual stock fluctuations, and the fine-tuning of take-profit and stop-loss conditions may all affect the subsequent stock trading sequence to undergo great changes (this is normal). The impact is a bit too large (sometimes a year’s return will change by more than 10 points). Although, from the multi-year backtest, the return drift is still within an acceptable range. (This year is more, that year is less, the overall return drift is not too outrageous.) However, this short-term return drift has not been thoroughly studied yet, and no method has been found to properly control it.

* There are still some details of the model that may need to be optimized, including buying timing, position management, crash response, of course, stock selection system, take profit and stop loss mechanism, etc.

* I used “Xiaomi plus rifle” to develop this short-term quantitative model: ordinary programmers + servers with general configuration + me, a long-term lazy person, to build the overall investment framework. From an external point of view, the probability of being able to polish a “short-term money-making artifact” in two years should not be high. In case this model fails after verification, just play it as an adult game.

* This short-term quantitative model will use a lot of short-term technical indicators, which is something that a price investment player like me has always resisted or even despised, but I still have an open mind to learn and apply it (I engage in value speculation, and I have my own way. sets, do not look at technical indicators at all). In addition, I used to think that it would be very schizophrenic to engage in long-term price investment and short-term technical operations at the same time. For example, on the day when a certain West Bank fell sharply, this short-term model suggested that Shenglu Communications had stopped profit ahead of time (with a sharp drop in profit remaining 5%+) and Jiangzhong Pharmaceutical’s profit was only 3%+ (if it does not sell, there will be a temporary 10%+ floating profit), but my existing long-term model continues to buy Qianjin Pharmaceutical, Tianjian Group, etc. on the same day. On the surface, the buying and selling are reversed, which is indeed a bit of a split, but I agree with the statement that “human judgment can never catch up with a simple mathematical model”, and I invest according to the modularization of the model and “ruthlessly” implement the predetermined rules without any entanglement. . Of course, this is just an investment experiment.

The above investment thinking, I hope to inspire others when they build their own investment models (regardless of long and short), and hope to get corrections and constructive discussions.

@Today’s topic

#financial geek investment thinking#

$Jiangzhong Pharmaceutical(SH600750)$ $Shenglu Communication(SZ002446)$ $Qianjin Pharmaceutical(SH600479)$

This topic has 20 discussions in Snowball, click to view.
Snowball is an investor’s social network, and smart investors are here.
Click to download Snowball mobile client ]]>

This article is reproduced from:
This site is for inclusion only, and the copyright belongs to the original author.

Leave a Comment