 Welcome to this presentation about my paper Chaos Measure Dynamics in a Multifactor Model for Financial Market Predictions. This paper is the seventh paper of a five year long research study and represents its respective conclusion. The core finding of the research study is that chaos dominates financial markets. However, I will provide you with the initial situation and past research findings first Before elaborating on the current study's implications, first, we analyzed daily cascadic wavelet-denoist S&P 500 logarithmic returns of the past 20 years and have shown the underlying dynamics to resemble a mixture between deterministic hyperchaotic and quasi-periodic laminar stochastic dynamics. By application of a novel self-developed framework, I could discern the strange fractal attractor reconstruction of the S&P 500 return system and show inherent multi-fractal power law distributed scaling laws. Second, via empirical experiments, we found that the Hearst exponent represents fractal trending characteristics instead of long memory, if exceeding the value of 0.5 and is thereby implying the momentum effect. Moreover, the Hearst exponent has been shown to be time-varying and yielding own dynamical emotions. This implies that A. Mandelbrot's initial interpretation of the Hearst exponent was wrong and B. That studies using this insight to show long memory potentially stated multi-fractality instead. Long memory effects exist in parallel and are indicated by significant slow decaying auto-correlation functions instead. Finally, based on the denoted findings, the efficient market hypothesis is totally invalid in any variation. Next, I want to provide a bit of context found during analyzing thousands of papers on how chaos, multi-fractals and the Hearst exponent harmonize within financial markets. A dissipative dynamical system is a system with a significant and positive maximum Lyonpunov exponent and a negative Lyonpunov exponent spectrum sum, which are both given for the S&P 500 return system. Dissipative systems collapse onto their own strange fractal attractors, which can then be intersected by Poincaré sections and result in fractal set representations. These fractal sets can be analyzed and described by multi-fractal analyses and power law scaling distributions. These fractal characteristics are indicated by the Hearst exponent, which is demonstrated in the literature to be time-varying. This implies that fractality itself is time-varying, thus the system attractor, i.e. the chaotic motion of the trajectories are also varying over time. This raises several complications and research questions, since it is unclear a. if chaos is time-varying and b. if there are given laws of motion for these time-varying chaotic dynamics. Furthermore, there is a notable gap in the literature on how to quantify such chaotic variations over time. To provide a solution, the presented paper applies three different sized scrolling windows to chaos measures, i.e. the maximally ampoune of exponent as measure of exponential divergence rate of system trajectories, the referringly ampoune of exponent spectrum sum, as indication of the system's underlying nature, the sample entropy as measure of information content or fractality inducing self-similarity, and finally the Hearst exponent, which provides information about the system's nature, i.e. valid efficient market hypothesis or Brownian motion case, fractality or mean reversion. As a result, I could show that these arising chaos measure series are time-varying, i.e. indicating unstable or time-varying chaotic motions of the S&P 500 return system. Further, the time-varying chaos measure series, or CMS for short, are shown to be co-integrated with the original S&P 500 price series by applying a vacuum Johansson test approach. Next, I showed that the S&P 500 prices could be modeled by a deep-learning multiperceptron regressor dynamic multi-factor model with the CMS as feature inputs. The approach could be successfully verified by stating IID residuals. Finally, by applying the novel framework, I could identify the empirical data-generating process characteristics of the CMS, i.e. quantifying the chaotic time-variation of the underlying system. In short, I will show the graphics of the CMS. First, we see the Hearst exponent time-variation. The red line indicates the validity of the efficient market hypotheses, i.e. the case of a Brownian motion. Below, the red line is mean reversion, while above the red line indicates fractality, i.e. trending characteristics. The blue bars indicate crisis periods. We can see that, before or after crisis periods, a shift from strong trending to mean reversion is visible, thus explicating vanishing momentum. The vanishing momentum effect is the cause for so-called momentum crashes, which can be directly observed regarding time-varying Hearst exponents. Next, we see the maximum Lyampunov exponent dynamics. We see that shifts from positive to negative exponents and a large volatility is present. To continue, we see the Lyampunov exponent spectrum sum, which also indicates a shift between conservative and dissipative system structures, while yielding large regimes of high volatility. Finally, we see the sample entropy variations, which imply a surge before crisis periods, i.e. an increase in fractality or information density. To continue, I apply the shown CMS into a vector error correction model, or VACM for short, combined with a Johansson test approach to validate potential relations, correlations or better called co-movements in terms of co-integrations. I selected a closed system view, since I only apply the CMS originating from the original system for it. We can see that the VACM is not suited for the relations, however, the eigenvalues are significant, indicating that co-integration is present. This means we are in need for a better model specification. Therefore, I implemented a deep learning multi-perceptron regressor as dynamic multi-factor model representation with four of the CMS as factor inputs. The model is validated by analyzing the residuals, resulting in IID residuals. However, slight leptochartic tails are found in the distribution, which is common in financial markets data. Regarding the performance metrics of a backtest with around 1250 steps, we can see that the model is acceptable. However, for real world trading, the model is not detailed enough and maybe is required to be further specified as indicated by the given performance metrics. The denoted backtest can be seen visualized here. In summary, I can state that dynamic chaotic motions can model financial markets. I showed that time-varying CMS imply chaos instability and time variation for the underlying dynamical system. Further, the CMS are co-integrated. The CMS are shown to be factors for a dynamic multi-factor model based deep learning multi-perceptron regressor under a closed system view. This means performance can drastically increase by including exogenous variables such as macroeconomic indicators etc. Finally, quantifying the empirical DGP of the CMS state them to yield own dynamical motions which are different from the originating system. To conclude, I can state that the exploitation of the underlying empirical DGP via the CMS information can be used to potentially generate alpha-generating investment strategies. Furthermore, I hypothesize that financial markets are solvable. Finally, I want to state that the presented approach is applicable to any kind of non-linear time series thus be generalizable in this domain. This concludes my presentation and I will answer questions and remarks.