BucketofBalls99 said:
My bad. Was catching up on some things and totally missed it
Stock Markets
25,802,420 Views |
235175 Replies |
Last: 5 hrs ago by aggies4life
AMD near $100, I'm a buyer
Curious your thoughts on AMD. I'm too emotionally invested in this one. My avg is now 119 after a double up ~138. My first individual stock purchase and I love the CEO. Just curious what you're seeing.
Curious your thoughts on AMD. I'm too emotionally invested in this one. My avg is now 119 after a double up ~138. My first individual stock purchase and I love the CEO. Just curious what you're seeing.
It's a solid company and my call is only based on it's charts atm. It confirmed a 'death cross' the 3rd week of November and has set new 52 wk lows recently because it's buyers' confidence has been shaken. Fundamentally, they're a solid winner, but short term, they're vulnerable to retest stronger support near $100.Quacked said:
AMD near $100, I'm a buyer
Curious your thoughts on AMD. I'm too emotionally invested in this one. My avg is now 119 after a double up ~138. My first individual stock purchase and I love the CEO. Just curious what you're seeing.
For me, $100 is highest target for re-entry. I'd prefer low $80's though..Quacked said:
AMD near $100, I'm a buyer
Curious your thoughts on AMD. I'm too emotionally invested in this one. My avg is now 119 after a double up ~138. My first individual stock purchase and I love the CEO. Just curious what you're seeing.
I can also AMD testing that range if we have a violent correction but I'll start buying at $100 and add more if it continues to $80.
$WWR
0.811
+0.231 (39.78%)
Dec 27, 2024, 3:16 PM EST - Market open
0.811
+0.231 (39.78%)
Dec 27, 2024, 3:16 PM EST - Market open
The executive chairman of WWR is also the executive chairman of OPTT. In November he bought 100,000 shares at .148. The next week it started it's rally to .50 then tanked. Right before Christmas they issued another press release and stock resumed back up.
Pre ordered Lamborghini with a WWR license plate.
Well I ordered WWR graphite for the battery in my electric Lamborghini. Booyah.TxAG#2011 said:
Pre ordered Lamborghini with a WWR license plate.
On Lincoln Park if I understand the deal, Lincoln park can purchase shares at 97% of the lowest price of that day or the average of the the 3 lowest closing price from they 10 prior days before purchasing. There is no requirement to hold the shares.
The deal basically incentivizes a pump and dump. They are guaranteed a low price and can sell anytime. If you go back and look at the chart there are 3 day period where volume is high and the range is big. I'll have to go through the 10K to see if those are days shares were issued but I assume so. Also it's funny they have in the agreement they can't go above 9.99% of that shares. Basically Lincoln Park as a pass through to fund the company from retail shareholders. They aren't putting their money in this.
I looked into several deals Lincoln Park is part of and it's a lot of stocks that have been talked about here
WWR
IBIO
MMAT
HUMA
MVIS
PHUN
WKHS
HUMA could be a nice short. I know it's popped recently.
The deal basically incentivizes a pump and dump. They are guaranteed a low price and can sell anytime. If you go back and look at the chart there are 3 day period where volume is high and the range is big. I'll have to go through the 10K to see if those are days shares were issued but I assume so. Also it's funny they have in the agreement they can't go above 9.99% of that shares. Basically Lincoln Park as a pass through to fund the company from retail shareholders. They aren't putting their money in this.
I looked into several deals Lincoln Park is part of and it's a lot of stocks that have been talked about here
WWR
IBIO
MMAT
HUMA
MVIS
PHUN
WKHS
HUMA could be a nice short. I know it's popped recently.
Here's a transcript of part of a story with NPR from this past summer. What I found interesting was that the proprietary process WWR had for refining was basically crap. Nobody wanted it. That was surprising because I remember OA posting once about talking to the engineers and they were giddy. It turned out whatever they had was crap but they've figured it out now they say and I guess the proof is the offtake agreements. That help explains why it took so long to get financing or any deals.
JON JACOBS: So what you're looking at here is the world's largest blender.
HELM: We're looking down into this big, circular machine. It has blades that'll mill the graphite to a smaller size because graphite can't just come out of the ground and go straight into a battery. It has to be coated and reshaped.
JACOBS: Right now we have 10 spheroidizing machines installed. And so this would be the next step.
HELM: Spheroidizing. To work in batteries, graphite particles need to be rounded, kind of potato-shaped. You can think of this as problem No. 1 for getting a battery industry off the ground in the U.S. - the potato problem. Chinese companies already know how to reshape graphite. But when Wes****er Resources first tried it, their particles were too flat.
JACOBS: The feedback months later, through our battery partners, would be, make it more of a sphere. So we'd go back to them again, and they would do it again.
HELM: Do you get sick of hearing that - like, more of a sphere?
JACOBS: Yeah, you do. Yeah. Believe me. Yeah. It just never seems to be spheroidal enough to...
HELM: He says they've now figured it out. But if the U.S. wants to make batteries here, this is the kind of technical know-how that stands in the way. Another thing a new industry needs is new infrastructure. In this case, to get up and running, the plant needed a new sewer pipe for their wastewater. That's problem No. 2 - the pipe. And I heard about that from Woody Baird. He's the mayor of nearby Alexander City, Alex City to locals. The graphite plant would mean good-paying jobs for people in town, and so Mayor Baird took out a big loan to build the pip
JON JACOBS: So what you're looking at here is the world's largest blender.
HELM: We're looking down into this big, circular machine. It has blades that'll mill the graphite to a smaller size because graphite can't just come out of the ground and go straight into a battery. It has to be coated and reshaped.
JACOBS: Right now we have 10 spheroidizing machines installed. And so this would be the next step.
HELM: Spheroidizing. To work in batteries, graphite particles need to be rounded, kind of potato-shaped. You can think of this as problem No. 1 for getting a battery industry off the ground in the U.S. - the potato problem. Chinese companies already know how to reshape graphite. But when Wes****er Resources first tried it, their particles were too flat.
JACOBS: The feedback months later, through our battery partners, would be, make it more of a sphere. So we'd go back to them again, and they would do it again.
HELM: Do you get sick of hearing that - like, more of a sphere?
JACOBS: Yeah, you do. Yeah. Believe me. Yeah. It just never seems to be spheroidal enough to...
HELM: He says they've now figured it out. But if the U.S. wants to make batteries here, this is the kind of technical know-how that stands in the way. Another thing a new industry needs is new infrastructure. In this case, to get up and running, the plant needed a new sewer pipe for their wastewater. That's problem No. 2 - the pipe. And I heard about that from Woody Baird. He's the mayor of nearby Alexander City, Alex City to locals. The graphite plant would mean good-paying jobs for people in town, and so Mayor Baird took out a big loan to build the pip
Well I guess it's time to steal Chinese graphite refining tech
WWR team probably stalled out when no Hunter paintings were offered for federal grants/support. Biden wouldn't ever think about encroaching on Xi's graphite market share. Now, we got tarrifs on table and Musk who has been urging domestic battery mining for years. No idea if Gov Ivey has an open line to WH, but if she did, I would think this venture would be a topic.
Will be interesting to see how Tesla holds up this week with Elon's war with MAGA heating up on Twitter.
Merry Christmas and Happy New Year to everyone on this fantastic board.
So I've done a thing! Or, at least I think I have.
The past few weeks I've been trying to dig into machine learning and advanced predicting algorithms. Since Trump is taking office in January, I figured I would start by taking stock history for all the stocks in the NYSE and NASDAQ since 2016 and use it to train my machine learning model. The shear number of stocks involved forced me to filter out any stock that traded below 1 million shares a day and under 1 dollar a share. This put me around 1250 stocks in total to look at. Maybe when I get home and onto my PC, I can start relaxing this requirement a bit, but at the moment, this is the best my MacBook can handle.
The model uses the Random Forest alogorithm mixed in with LSBoost for gradient boosting. I'm hardly an expert at all this, but this is what chatgpt suggested I use. Random Forest is used to find broad patterns, while LSBoost uses iterative learning for detailed refinement.
I'm using 9 indicators at the moment to train my model.
I'm considering Bollinger Bands in next iteration and eventually, I'd like to do some monte carlos based on lines of support/resistance....but baby steps.
When Random Forest runs, it identifies what the key inputs to its learning were. Generally speaking, you don't want a single variable to dominate in the model and you want to avoid giving anything directly related to price to avoid over fitting. Random Forest is apparently better at not overfitting than other codes, but you still have to be diligent. In my case, instead of raw emas, I normalized them first to obfuscate the raw price from the model. Below is how the features broke down and how relevant they were to the predictions.
Part of the process with Random Forest is setting the number of trees the model uses to learn and tracking the out of bag error after each iteration. For me, my OOB error started off pretty low and trended down with each run. I think I was using something like 100 trees in this version. This part makes me a little weary in that it seems too good to be true. I would feel better about it if I had more error actually as this amount of perfection is a sign of over fitting.
Then the moment of truth....how well does the model predict what actually happened?
The blue is the historical data. The red is what the model is predicting will happen. It's cheating because it's using the known price to determine the the variables and then using the variables to predict the price. Still, it shows that the model is able to predict the price with the indicators. The green curve shows where the model thinks the stock is headed over the next 40 days and is the true forecast. After each daily prediction, it updates the price and all the features that went into the original model. I'll say, it does a poor job of capturing the day to day variation and I'm not sure how I feel about that. I was hoping for something super fancy where it attempts to capture the ups and downs, but as it stands you really have to use it as a general trend tool and not so much day trading right now.
So how did the model do at predicting the price compared to actuals? Note, I did not feed it ANY 2024 data past January 4th so everything that happened this year is a completely authentic result. Here, I've sorted all 1250 stocks and by %return and am just showing the top 10.
If you blindly bought the stocks the code suggested, you would have ended up +122% for the year. Not all stocks hit the targets. Some of them hit within my 40 day prediction period. Some of them hit sometime after. None of them were losers. I got my computer churning away for 2025 now if you're interested. Might take a few more hours to run, but I was giddy this morning to see these results.
So I've done a thing! Or, at least I think I have.
The past few weeks I've been trying to dig into machine learning and advanced predicting algorithms. Since Trump is taking office in January, I figured I would start by taking stock history for all the stocks in the NYSE and NASDAQ since 2016 and use it to train my machine learning model. The shear number of stocks involved forced me to filter out any stock that traded below 1 million shares a day and under 1 dollar a share. This put me around 1250 stocks in total to look at. Maybe when I get home and onto my PC, I can start relaxing this requirement a bit, but at the moment, this is the best my MacBook can handle.
The model uses the Random Forest alogorithm mixed in with LSBoost for gradient boosting. I'm hardly an expert at all this, but this is what chatgpt suggested I use. Random Forest is used to find broad patterns, while LSBoost uses iterative learning for detailed refinement.
I'm using 9 indicators at the moment to train my model.
- ema5
- ema10
- Volatility
- RSI
- ema5 Gradients
- 3 Day trading range (high)
- 3 Day trading range (low)
- RSI Gradients
- Approximate Volatility
- RSI Divergence (Bullish)
- RSI Divergence (Bearish)
I'm considering Bollinger Bands in next iteration and eventually, I'd like to do some monte carlos based on lines of support/resistance....but baby steps.
When Random Forest runs, it identifies what the key inputs to its learning were. Generally speaking, you don't want a single variable to dominate in the model and you want to avoid giving anything directly related to price to avoid over fitting. Random Forest is apparently better at not overfitting than other codes, but you still have to be diligent. In my case, instead of raw emas, I normalized them first to obfuscate the raw price from the model. Below is how the features broke down and how relevant they were to the predictions.
Part of the process with Random Forest is setting the number of trees the model uses to learn and tracking the out of bag error after each iteration. For me, my OOB error started off pretty low and trended down with each run. I think I was using something like 100 trees in this version. This part makes me a little weary in that it seems too good to be true. I would feel better about it if I had more error actually as this amount of perfection is a sign of over fitting.
Then the moment of truth....how well does the model predict what actually happened?
The blue is the historical data. The red is what the model is predicting will happen. It's cheating because it's using the known price to determine the the variables and then using the variables to predict the price. Still, it shows that the model is able to predict the price with the indicators. The green curve shows where the model thinks the stock is headed over the next 40 days and is the true forecast. After each daily prediction, it updates the price and all the features that went into the original model. I'll say, it does a poor job of capturing the day to day variation and I'm not sure how I feel about that. I was hoping for something super fancy where it attempts to capture the ups and downs, but as it stands you really have to use it as a general trend tool and not so much day trading right now.
So how did the model do at predicting the price compared to actuals? Note, I did not feed it ANY 2024 data past January 4th so everything that happened this year is a completely authentic result. Here, I've sorted all 1250 stocks and by %return and am just showing the top 10.
If you blindly bought the stocks the code suggested, you would have ended up +122% for the year. Not all stocks hit the targets. Some of them hit within my 40 day prediction period. Some of them hit sometime after. None of them were losers. I got my computer churning away for 2025 now if you're interested. Might take a few more hours to run, but I was giddy this morning to see these results.
Very interested! Really good stuff
Interesting. I suppose checking last year's numbers first will guard against hallucinations for next year. Please post some of its recommendations so we can look back next December at how it did.
Trump will fix it.
Had trouble getting that table to load. See my previous post now because that's exactly what I did. Will post 2025 when the model finishes this afternoon
In for the 2025 AI recommends.
Looking forward to its recommendations. Then checking back in March like in your table. The only thing I see causing a drastic difference would be unforeseen circumstances. Declaration of war, natural disasters, etc. Common risks I suppose.
This is extremely interesting. Thanks for posting it.
This is extremely interesting. Thanks for posting it.
Trump will fix it.
Awesome! Look forward to the 2025 picks!
This is rad
Pretty dang cool, but again, seems like there's a massive feedback loop there. Testing in on the unknown of the future will tell the tale. Still, pretty unique. I appreciate your hard work and look forward to more.
I will say, I simply don't think <10 years is enough data. Everything that happened in tha timeframe was in the context of an ongoing 14 year bull market. 2020 was a minor blip of bearishness that was gone as fast as it came. 2022 the same. I'd like to see you test it going back to the 70's, but I know that's a ton of work and severely limits the amount of stocks included.
I will say, I simply don't think <10 years is enough data. Everything that happened in tha timeframe was in the context of an ongoing 14 year bull market. 2020 was a minor blip of bearishness that was gone as fast as it came. 2022 the same. I'd like to see you test it going back to the 70's, but I know that's a ton of work and severely limits the amount of stocks included.
Going all in with the 25 picks! Ride or die?
Count me in.
Agreed about the feedback loop when it comes to the backtesting plot (with the red/blue/green), but when it comes to the table I provided, there is exactly zero feedback going into it. It's 100% an authentic prediction along with the result.
I can definitely try to get back to 1970s data. I think the bigger challenge (along with the pure amount of data involved) is going to be finding stock tickers that existed back then that are still around today. Pretty good chance they've changed their names at least once since then which may result in less useful data that we'd like. Still, something to add to my list.
As far as 2025 goes, here's what the model shows me today. I've done exactly ZERO technical analysis on these, so take it with a massive grain of salt. I think it's interesting that SAVA made the list because I definitely had no intention of buying them at this point. In any case, here is my models top 20 pick for the next couple months. I plan on tweaking the model as we go, but I'll consider this 2025s starting point.
I can definitely try to get back to 1970s data. I think the bigger challenge (along with the pure amount of data involved) is going to be finding stock tickers that existed back then that are still around today. Pretty good chance they've changed their names at least once since then which may result in less useful data that we'd like. Still, something to add to my list.
As far as 2025 goes, here's what the model shows me today. I've done exactly ZERO technical analysis on these, so take it with a massive grain of salt. I think it's interesting that SAVA made the list because I definitely had no intention of buying them at this point. In any case, here is my models top 20 pick for the next couple months. I plan on tweaking the model as we go, but I'll consider this 2025s starting point.
Bookmarked. Thanks!
Trump will fix it.
Which one of you is going to buy every single one of these tomorrow? We need a hero (guinea pig).
Other than SAVA I like the top half of that list. I am game.
I don't think anyone is sighing up for a sunday trade.Heineken-Ashi said:
Which one of you is going to buy every single one of these tomorrow? We need a hero (guinea pig).
Here's one - FDX.
Such a hard chart to get a grasp of. Seriously. Look at this..
The last 5 earnings has either seen it rocket up $50, down $50, and sometimes both. Since October 2023 it's been within a $100 range chopping bottom to top and back down. I believe this is an ending diagonal. Here's how I have it charted.
And I think it gets one more rise aiming for $330-$350. Could happen before February earnings, or could be like the last 5 times and happen on it. Could also spike up and immediately reverse before you can blink.
In the short term, it could test the lower trendline one more time which is a $20 drop from current price. So if you can't stomach a $20 per share loss, don't play. But if you can, it's $20 risk to $50-$70 reward.
In the long term beyond earnings, I'd be out and not even look at it again. If I'm right and it hits that range, the next trip will be down to $210-$225, and possibly never seeing that high again.
Such a hard chart to get a grasp of. Seriously. Look at this..
The last 5 earnings has either seen it rocket up $50, down $50, and sometimes both. Since October 2023 it's been within a $100 range chopping bottom to top and back down. I believe this is an ending diagonal. Here's how I have it charted.
And I think it gets one more rise aiming for $330-$350. Could happen before February earnings, or could be like the last 5 times and happen on it. Could also spike up and immediately reverse before you can blink.
In the short term, it could test the lower trendline one more time which is a $20 drop from current price. So if you can't stomach a $20 per share loss, don't play. But if you can, it's $20 risk to $50-$70 reward.
In the long term beyond earnings, I'd be out and not even look at it again. If I'm right and it hits that range, the next trip will be down to $210-$225, and possibly never seeing that high again.
Looks like you basically have 1 day to take advantage of it?
It's the holidays and I have a baby on the way Monday. My brain can't track what day it is right now lol.Philip J Fry said:I don't think anyone is sighing up for a sunday trade.Heineken-Ashi said:
Which one of you is going to buy every single one of these tomorrow? We need a hero (guinea pig).
P.S., I won't be around until mid-week at the earliest unless I have some time to look on my phone. Probably won't be very useful. Yall dont break things. My last baby was born a week before COVID. If we get a crash next two weeks, the Heineken bearish indicator warned you.
Best wishes for a safe delivery of Heine-Lite. Please let us know how things are going.
Best wishes for a safe delivery!
Featured Stories
See All
Scouting Report: No. 10 Texas A&M at No. 17 Oklahoma
by Tom Schuberth
A&M's growing 2026 class features pledges from across the country
by Ryan Brauninger
8:10
12h ago
909
7:00
12h ago
3.7k
Members of A&M's 2025 signing class officially make Aggieland home
by Ryan Brauninger