Capitalico New Release: Automatic Exit Rule Optimization

Capitalico is newly equipped with our Automatic Exit Rule Optimization function. With this update, we offer you the option of a fully automated exit rule strategy that improves algorithm performance and frees your time to focus on other aspects of your trades.

Getting Even Smarter

Capitalico brings the latest AI technology to your investment trading algorithms.  

Automatic Exit Rule Optimization is another advancement in how Capitalico combines personalization and automation to your trading investments. By automatically selecting the most profitable ratio of loss cut/profit take from any selected backtesting period, Capitalico frees you from the tedious task of analyzing years of backtesting data.

With a click of the checkbox, you can activate Capitalico’s Automatic Exit Rule Optimization function and experience how AI technology can work for you!

スクリーンショット 2016-04-28 10.18.57

Adding Ichimoku Cloud

This update also includes Ichimoku Cloud indicators to display on your trading chart. Include this in your data sets to have it learned as a part of your algorithmic training data.

As always, we appreciate your interest in Capitalico, and please do not hesitate to contact us for any questions you may have.




Capitalico is coming back to the web browser!

Screen Shot 2016-04-18 at 11.39.15 PM

Dear friends,

I’m excited to inform you that Capitalico is moving back to the web platform, effective today!  Now you can build your trading algos through your desktop and laptop browsers without an iOS device. The Capitalico mobile app will still help you with your trades, and starting with the next version, it will function as a viewer of your web-based algo builds and receiver of live notifications.

After the initial launch of Capitalico, we have spent a lot of time interacting with our users and have worked to incorporate all the great suggestions. The most important of which is that we realized the browser interface is the best option to look intensively at the technical charts and customize builds. Interestingly, we had already implemented the beta version of Capitalico on a web platform late last year, but moved towards mobile in efforts to simplify the process and make it more accessible to everyone. The decision to make Capitalico browser based has been difficult but well researched. Having beta tested both versions, we asked ourselves, “which platform would be more suitable for Capitalico?” and it became clear from the users’ voices that the web browser interface is the right way to go.

Capitalico is still in early development but with every step, we make progress and build our user base. We appreciate everyone, and thank you for your valuable feedback. Our users play a crucial role in how we shape Capitalico, and your voice is important to us.

Thank you,

P.S. as part of today’s release and per user request, Capitalico is adding another currency pair, USD/MXN (editable only on the web browser version).  We will continue to add more assets and indicators, as well as different ways of helping your trading life.  More to come later, so please stay tuned!


Hitoshi presented our deep learning technology at GTC 2016

On Apr 6th, 2016, Hitoshi, CTO of Alpaca, talked about our deep learning technology around chart pattern recognition using LSTM at NVIDIA’s GTC (GPU Technology Conference) 2016, as well as TensorFlow Mega Meetup on Apr 3rd.  The slides are uploaded to SlideShare.

Capitalico / Chart Pattern Matching in Financial Trading Using RNN from Alpaca

The session was full with some people standing on wall.  Capitalico is one of the most advanced services that utilize the cutting-edge deep learning technology and we shared lots of idea that many other projects can benefit from.  We also recognized financial technology is one of the most interesting space where machine learning and deep learning can solve hard problems.  We continue to advance our technology, and thank you all who came to the session!!

Launching Capitalico App on AppStore!!

Alpaca: Deep-Learning FinTech Startup Launches Capitalico App for Anyone to Make Trade Ideas into AI Algorithms

No computer programming required to build sophisticated investment trading algorithms

March 11, 2016 (San Mateo, CA) 

Alpaca, the leading startup in Deep-Learning FinTech launched the anticipated, personal investing application “Capitalico” today on the Apple App Store for mobile users. Capitalico is a Deep-Learning trading platform that enables traders to automate their trade ideas without any programming or market investing experience, and has been running after a closed beta launch to limited, invite-only users for last several months. Today, Capitalico is being launched to the public for iOS users and available as an iPhone application.



Alpaca fully leverages its Deep-Learning and Parallel Distributed technologies built by industry veterans in application development from companies such as EMC/Pivotal, Greenplum, Microsoft, and Nokia. Capitalico distinguishes itself by building its own unique technology based on financial time series data and user specifications. With the Capitalico application, Alpaca empowers traders to automate their trade ideas into AI algorithms with only few taps on their mobile devices.


“We saw an opportunity to close the gap that most traders have a difficult time to verbalize their trading rules precisely enough to be coded as automated trading algorithms,” said Yoshi Yokokawa, Alpaca’s co-founder and CEO. “Capitalico is going to support many traders’ daily activities by enabling them to build their customised trade notifications and automated trading algorithms through highlighting their winning entry patterns on Capitalico app.”


Capitalico is now publicly available on the Apple App Store for free. The app is available worldwide in English and Japanese on the Apple App Store. Please visit for more Capitalico information including videos and screenshots of the app.


To download the app:


About Alpaca

Alpaca is a VC-backed Silicon Valley Fintech startup that helps individual traders build automated trading algorithms using Deep-Learning, to expand platform to be an algo-trading marketplace eventually. The first public mobile app “Capitalico” is publicly available now on the Apple App Store, and is supported by 10+ year veterans from AI, database For more information, visit

Capitalico Development Update #003

This is Yoshi from Alpaca. I would like to update you regarding Capitalico development which I sent an email to those who are on the Capitalico waiting list (you can sign up from Capitalico waiting list).

In this update, I would like to explain where Capitalico is headed to in the long term, and share with you about the short and mid term development plan.

Where Capitalico is headed to in the long term

We are planning Capitalico to be an online marketplace for the investment products that directly connects two types of individuals who build the products and who buy the products.

We think that there is a continuous trend of automating the transactions and asset-allocation when making investment products. We also think that the algorithms, which automate such investment decisions, will take even bigger portion in investing in general. Considering these trends, we are releasing Capitalico mobile application to set up the infrastructure for individuals, where they are empowered to build automated investment algorithms to enter the trend of automated investing.

Short/Mid term development roadmap for Capitalico

For Capitalico v1.0 release at this February-end, we focused our development effort so that users can build their own investment algorithms on a mobile-app without needing much professional knowledge. Because of this, we decided to keep the functions minimally enough to let little-experienced users build algorithms. In the short term, we are going to maintain this focus while setting below points as our high priority.

*Stock market integration

Based on the initial user feedback, we think that the stock market integration is one of the most important milestones for Capitalico. We are planning to launch this major update where you can build the algorithms to trade in stock market. We are targeting this update in 2016Q2.

*Fully automated algorithmic trading

We are in process of integrating Capitalico app with a few brokerage firms to enable you to trade fully automatically by using the algorithms that they build. We plan to announce these brokerage partners during 2016Q2. Before the release of  the automated algorithmic trading feature, you can make trades based on the real-time trade signals that are sent out from the Capitalico application through iPhone push notifications and/or email notifications.

*Android version

We think that it is important for Capitalico to be available on the Android devices as well. At this February-end V1.0 release, Capitalico will be available only on iOS devices. However, we plan to release the Android version during the 2016Q4 after releasing the major update of stock market integration.

*Enhancing the total experience of Capitalico

Although I mentioned about “keep the functions minimally enough to let little-experienced users can build investment algorithm,” we are fully aware of the strong needs of enhancing the total experience and function of Capitalico. Through these enhancements, we want to make sure that well-experienced users are also satisfied with using Capitalico to build highly sophisticated investment algorithms. We are carefully deciding the timeline of these enhancements listed below to make sure these happen sooner than later.

  • Add more types of indicators available
  • Add more currency pairs available
  • Support multi-chart pattern matching feature
  • Algorithm transfer feature among different timeframes and asset-classes
  • Enhancing the performance analytics feature
  • Enhancing the exit rule setting
  • Adding the trend-line/channel function
  • Adding the money management setting
  • Releasing the PC browser version

I am planning to send out the next update in the week of 2/8 where I will be announcing the new Capitalico landing page.

As always, we appreciate your interest in Capitalico, and please do not hesitate to contact me for any questions or inquiries that you may have.

Best regards,


Hitoshi speaks at NVIDIA’s GTC in San Jose this April

Alpaca’s CTO, Hitoshi, will be speaking at NVIDIA’s GPU Technology Conference held in San Jose this April. Hitoshi plans to talk about the deep-learning technology that is used in Capitalico application. Come join his talk at the GTC!!

S6309 – Capitalico – Chart Pattern Matching in Financial Trading Using RNN

Hitoshi Harada CTO, Alpaca

Discretionary trading by technical analysis and momentum strategy in the financial market has been difficult to automate by quant-style rigid conditional programming as it involves a lot of fuzziness and subtleties of human perception. Our application, Capitalico, analyzes the financial time-series data and trader’s behavior to solve this problem using the RNN/LSTM. In this talk, we’ll introduce the problem and our approach, and detail pitfalls and practices, such as how we choose networks and parameters to achieve the best accuracy and performance with deep learning using GPUs. As we borrowed great ideas from past deep learning applications, we’ll help you understand how we converted those ideas to our solution and how to apply deep learning to your problem.

Level: Intermediate

Type: Talk

Tags: Finance; Deep Learning & Artificial Intelligence



Capitalico Development Update #002

Dear friends,

This is Yoshi from Alpaca. Today, I would like to update you regarding Capitalico development and what you should expect from the public beta launch targeted at this February-end (“v1.0”).



<Capitalico development update as a mobile-first app>

In last update, I let you know that we decided to make Capitalico a mobile-first app. We fixed the design that is going to be used for the public beta launch, and are now working on implementing the design. We are continuously working to improve the accuracy as well as the speed of the underlying deep neural nets.



<What pain Capitalico v1.0 is going to solve>

I want to clarify what pain Capitalico v1.0 is going to solve at this public beta release in February-end. We have an exciting development plan ahead, but it takes a while for us to get to the ultimate version of Capitalico.

We are targeting mainly three user groups in v1.0.

(1) Forex discretionary traders often have a difficult time making new trading strategies or follow existing strategies, because they manually need to watch multiple charts and go back the time historically, which takes a lot of effort. Capitalico solves this pain by providing them a simple mobile app that lets them automate their trade ideas without using any programming. This allows them to repeat the PDCA cycle of making new trading strategies, and supports them to follow the existing strategies by sending them real time trade signals.

(2) Most forex systematic traders are not programming experts, therefore most cases, they need to learn programming or outsource the programming task when they want to make automated trading algorithms. MetaTrader is a famous tool to write programming to make automated trading algorithms. However, it is no use if a user cannot write programming. They often purchase pre-made trading algorithms or subscribe to the third-party trading algorithms. Capitalico solves this pain by letting them make their own automated trading algorithms without using any programming or programming-like logics, but only through interaction on charts. They do not need to buy algorithms from other people, but now can make as many algorithms as they want on Capitalico.

(3) Many people who are interested in market and algorithm trading have a difficult time entering the market, because there is no appropriate tool for them to try out to see how their trade ideas would work unless they can write programming. Capitalico solves this pain by providing a mobile app that allows them to realize their trade ideas into algorithms with several taps on a smartphone app. In addition, they can always track how their algorithms are performing in real market, so that they can see how well their ideas work.

If you are familiar with MetaTrader software, you can think of Capitalico v1.0 as programming-less version of MetaTrader, that even works as an independent mobile application.


<What you can do on Capitalico v1.0>

For Capitalico v1.0 public beta release, we focused our development effort so that we can minimally achieve to allow users to build their own investment algorithms on a mobile-app without much professional knowledge. I am listing up app specifications here.

  • Asset class – 7 major currency pairs (USD/JPY, EUR/USD, EUR/JPY, GBP/USD, NZD/USD, AUD/USD, USD/CHF)
  • Entry rule setting – Selecting specific chart patterns on single currency pairs’ single time-frame chart
  • Exit rule setting – Setting ratios of profit-take to stop-loss with automated optimisation feature
  • Capital management setting – N/A
  • Backtest – Backtesting against more than 10 years with 1-Minute chart
  • Livetest – Livetesting with 1-Minute chart and with Tick-chart for trade execution
  • For live-trade – mobile push notification and/or email notification for real time trade signals (some of this is a paid function)
  • Platform – iOS (iPhone)

We decided to cut out many features in order to release the minimum version of Capitalico for v1.0. However, we understand the strong needs from you for many features, functions, and asset classes, so we do have a plan already scheduled to implement them. For the detail product development roadmap and what Capitalico is going to be, I am planning to describe fully at the next development update, which I am planning to send out on the week of 1/26.

As always, we appreciate your interest in Capitalico, and please do not hesitate to contact me for any questions or inquiries that you may have.

Best regards,


New Chapter for Labellio

We are excited to announce that Labellio, the easiest deep-learning platform for computer vision, got a new owner for its future expansion, KYOCERA Communication Systems Co., Ltd. (“KCCS”).


It’s been half a year since we released Labellio as beta service last year, and we’ve been getting tons of positive feedbacks as well as various use-cases which we had never imagined. We see that there are now more services to access to the deep-learning technology, but we believe that Labellio is still one of the easiest and the most powerful custom image classification solution in the world.

In response to those strong demands, we, Alpaca as a startup, concluded that Labellio needs broader business backbone to go further. And this time, KCCS and Alpaca have agreed to hand over the comprehensive ownership of Labellio to KCCS effective on 19th January 2016.

Regarding the change of Labellio ownership, please take a look at KCCS’s press-release for more detail.

We continue to focus on FinTech sector by building Capitalico, “Build investment algorithms on Mobile.” We fully utilise our deep-learning technology to enable anyone to build complex investment algorithms with ease.


Capitalico Development Update #001


I would like to update you regarding Capitalico development which I sent an email to those who are on the Capitalico waiting list (you can sign up from Capitalico waiting list).  We are currently putting our best to deliver Capitalico to you as soon as we can.  To keep you in the loop, we would like to send you periodical updates.  It will not be too many, and we are thinking about one update for every month or every other month.

Oh, and I will remind you at the end of this email as well, but please be a part of Capitalico development effort by answering quick questionnaire here! 



The biggest update about Capitalico development is that we are going to release Capitalico as a mobile-first service.  Through feedback from our initial beta test users, we found out that many people do not have much time sitting in front of computer after coming home to play with Capitalico.  We are currently going through intensive design-sprints how we can make the right product as a mobile-first service.  We think that Capitalico can be the very first service that you can build trading algorithms on mobile comfortably, because it does not require you to write programming, especially in a small screen…


Release schedule

We are targeting the public release early next year… before the spring comes, and hopefully earlier than that.  However, even before the public release, we plan to open up the next beta versions to those who want early access to give us advise and feedback, so please stay tuned!


Want to hear from you!

It is super important for us to hear directly from you in developing Capitalico.  So we prepared 5-questions questionnaire form here, please let us know what you are thinking to be a part of the process!  


We appreciate your interest in Capitalico, and we do our best to deliver you something that makes you smile 🙂

How I built trading algorithm in a day without much trading experience

Building a trading algo is made easy by Capitalico.  I myself had very little experience in technical chart analysis, or to be honest I have never day-traded before, but was able to build profitable trading algo in a day so I’m explaining how.

The key indicator I’m using this time is Bollinger Band.  This indicator is to show moving average and moving standard deviation and looks like this.


There is a gray range around the candlesticks with a center line.  This center line is the moving average over the last time window and the upper and lower lines are the standard deviation over the same time window.  As you can see, the price movement tends to fit within this range, and that’s based on the standard distribution.  So on the flip side, you can also say that if the price goes out of range, it tends to come back closer to the center line.  Combining with two simple moving average lines, the price range is moving down but I thought at some point the price goes up again when it goes out of the band.  That is happening in the chart above, so I click the beginning of pattern and drag up to the entry point where price uptrend starts.  The right edge of the selection range should be the entry point where you buy this currency pair.

Then I click the “Make Algorithm” and wait for 30 minutes or so.  After it becomes ready for backtest, I try different set of backtest parameters.  It seems the confidence level of 0.85 should be enough to say the pattern is happening, and set “buy” with the loss cut/profit take at 10 pips.


Hmm, it’s not bad…  Keep in mind that it does not trade very frequently.  It is most likely because I have chosen only one range of the history, making the AI look for very strictly similar pattern to what I chose.  If you think other, slightly different patterns are acceptable, you are going to tell so by giving more of those patterns.

But it is still good that the strategy can win constantly.  Did I write any programming at all?  Note that this happened to be a good algo and many times you will see unstable or low performance ones, but that’s good as you can learn how it wins and loses and think again about your strategy, which is exactly what most professional traders are doing.

Hope this help you understand how you can build a trading algo with Capitalico.  We are running closed beta in a limited time and we value your feedback.  Sign up for the beta test today, and don’t miss this chance to get involved in the new trading era with Capitalico!

Enjoy happy trading life!