AWS Spot Termination - Is There a Better Way?

When AWS is running out of capacity, they give you a two minute notice and then they just terminate your instance. So what you want to do is you want to kind of prepare for that situation. What we get by Amazon is this, there's a strategy where you can set a baseline capacity that is on demand. But the problem here is if spot goes away, you don't grow over this 50% limit. And so how can I fall back to only mark capacity, which should be available, at least for a higher probability? We have an auto scaling group that runs a spot capacity and that auto scaling group is auto scale. That's our current situation.